Compare changes
Choose two branches to see what one has that the other does not, then open a pull request for it.
4 commits
- g1t is the stored author of what it opens; the person who asked is requested_by and keeps the author's rightsChase Pierceb1a555e
- g1t is one name: its agent's work, commits and comments show as @g1t, and nobody can claim g1t or g1t-agentChase Pierceaa2cb24
- Auto model routing: the cheapest tier that can do each piece of work, a retry goes up a tier, and each run records its tierChase Pierce07bae85
- One CI in the workflow list: hide files from before .github stopped being read, and gone files a live one replacedChase Pierceff663e5
| 75 | 75 | fi | |
| 76 | 76 | branch="runner-base/$(date -u +%Y%m%d-%H%M)" | |
| 77 | 77 | git config user.name "g1t" | |
| 78 | − | git config user.email "actions@g1t.sh" | |
| 78 | + | git config user.email "g1t@users.noreply.g1t.sh" | |
| 79 | 79 | git checkout -b "$branch" | |
| 80 | 80 | git add services/runner/base.json | |
| 81 | 81 | git commit -m "A new base image for g1t's sandboxes" |
| 14 | 14 | | --- | --- | | |
| 15 | 15 | | Something a person does on g1t.sh | The guide for it in `apps/docs/src/content/docs/guides/` | | |
| 16 | 16 | | An API route, a field, or an MCP tool | `apps/api/src/operations.rs` descriptions (they feed the OpenAPI document and the API reference), the example and notes for it in `apps/api/src/reference.json`, `reference/mcp.md`, and any guide that shows the call. Then refresh the docs' copy of the OpenAPI document with `G1T_WRITE_OPENAPI=1 cargo test -p g1t-api openapi`; `cargo test` fails until you do. | | |
| 17 | − | | How agents behave | `guides/g1t-agents.md`, and `apps/web/public/llms.txt` | | |
| 17 | + | | How agents behave | `guides/working-with-g1t.md`, and `apps/web/public/llms.txt` | | |
| 18 | 18 | | Settings, limits or prices | The page that names them, and the table it is in | | |
| 19 | 19 | | A new feature | A section in the guide that owns it, linked from the docs home if it is a new task | | |
| 20 | 20 |
| 316 | 316 | kind: PrincipalKind::Agent, | |
| 317 | 317 | acting: Some(Box::new(Acting { | |
| 318 | 318 | credential_id: "tok_1".to_owned(), | |
| 319 | − | agent: "g1t-agent".to_owned(), | |
| 319 | + | agent: "g1t".to_owned(), | |
| 320 | 320 | on_behalf_of: Principal::default(), | |
| 321 | 321 | scope: AgentScope { | |
| 322 | 322 | repo: RepoPath { |
| 268 | 268 | Op::UpdateIssue => "Update an issue", | |
| 269 | 269 | Op::CloseIssue => "Close an issue", | |
| 270 | 270 | Op::ReopenIssue => "Reopen an issue", | |
| 271 | − | Op::AssignIssue => "Assign an issue to the g1t agent", | |
| 271 | + | Op::AssignIssue => "Assign an issue to g1t", | |
| 272 | 272 | Op::Delegate => "Put an agent on it", | |
| 273 | 273 | Op::PlanWork => "Plan work", | |
| 274 | 274 | Op::GetPlan => "Get a plan", |
| 752 | 752 | "The check names reported on a repository's commits in the last 30 days, most recent first, with the events each was reported for: the names update_repo_settings takes in required_checks. A workflow's runs report a check named after the workflow; a check required on the default branch must be reported on a pull request's head (pull_request events) and, with the merge queue on, on its queued state (merge_group events)." | |
| 753 | 753 | } | |
| 754 | 754 | Op::MessageAgent => { | |
| 755 | − | "Send the agent working on a pull request a message: a correction, a hint, a change of plan. It receives it at its next step, and it is recorded in the pull request's session. The pull request's author, and anyone with the Write role or higher. An agent uses it to ask the agent on another pull request a question (kind: question) or hand it work that belongs there (kind: handoff), giving its own pull request as from_number; the answer comes back to it at its next step." | |
| 755 | + | "Send the agent working on a pull request a message: a correction, a hint, a change of plan. It receives it at its next step, and it is recorded in the pull request's session. The pull request's author (for one g1t made, whoever asked for it), and anyone with the Write role or higher. An agent uses it to ask the agent on another pull request a question (kind: question) or hand it work that belongs there (kind: handoff), giving its own pull request as from_number; the answer comes back to it at its next step." | |
| 756 | 756 | } | |
| 757 | 757 | Op::AnswerMessage => { | |
| 758 | 758 | "Answer a question or a handoff another agent sent you, by the message's id. For a handoff, set decline to say it is not yours to take. The answer reaches the asking agent at its next step." | |
| 791 | 791 | "Open an issue on a repository. Say what done means in the body if it helps, for instance under a \"Definition of done\" heading; what must pass before a pull request for it merges is the default branch's required checks, the same for every pull request." | |
| 792 | 792 | } | |
| 793 | 793 | Op::UpdateIssue => { | |
| 794 | − | "Change an issue's title, body, labels or the people it is assigned to. Only the fields given are changed; labels and assignees each replace the whole set. Its author may change their own issue; anyone else needs the Triage role or higher." | |
| 794 | + | "Change an issue's title, body, labels or the people it is assigned to. Only the fields given are changed; labels and assignees each replace the whole set. Its author may change their own issue, as may the person g1t filed one for; anyone else needs the Triage role or higher." | |
| 795 | 795 | } | |
| 796 | 796 | Op::CloseIssue => { | |
| 797 | − | "Close an issue without a pull request. Merging a pull request made for an issue closes it for you. Its author may close their own issue; anyone else needs the Triage role or higher." | |
| 797 | + | "Close an issue without a pull request. Merging a pull request made for an issue closes it for you. Its author may close their own issue, as may the person g1t filed one for; anyone else needs the Triage role or higher." | |
| 798 | 798 | } | |
| 799 | − | Op::ReopenIssue => "Reopen a closed issue. Its author may reopen their own issue; anyone else needs the Triage role or higher.", | |
| 799 | + | Op::ReopenIssue => "Reopen a closed issue. Its author may reopen their own issue, as may the person g1t filed one for; anyone else needs the Triage role or higher.", | |
| 800 | 800 | Op::PlanWork => { | |
| 801 | 801 | "Turn an outcome into a plan. An agent reads the repository and proposes the issues that would get there: what each changes, what done means for it (added to its body under \"Definition of done\"), the files it will touch, and which must merge before which. Returns the plan's id at once; the plan takes a minute or two to write, so read it with get_plan until its status is ready. Nothing is opened until apply_plan. Needs the Write role or higher." | |
| 802 | 802 | } | |
| 807 | 807 | "Open a plan's issues, each blocked by the ones it depends on. With assign, g1t agents start at once on every issue that depends on nothing, working in parallel, and on the others as what they depend on merges. keep limits it to some of the proposed issues, by their positions counting from 1. A plan is applied once. Needs the Write role or higher." | |
| 808 | 808 | } | |
| 809 | 809 | Op::AssignIssue => { | |
| 810 | − | "Assign an issue to the g1t agent. It opens a pull request for the issue in a sandbox of its own and sees it through: the repository's workflows run on it as its checks, a second agent reviews it, it revises when a check fails (reading the failing jobs' logs) or the review asks for changes, and it catches up when main moves. It is ready once the default branch's required checks pass and the review approves. Returns the pull request at once; follow its progress with get_pull_request. There is no model or agent count to choose. To put many agents to work, assign many issues. Needs the Write role or higher. In preview: only for accounts g1t agents are enabled for." | |
| 810 | + | "Assign an issue to g1t. It opens a pull request for the issue in a sandbox of its own and sees it through: the repository's workflows run on it as its checks, a second agent reviews it, it revises when a check fails (reading the failing jobs' logs) or the review asks for changes, and it catches up when main moves. It is ready once the default branch's required checks pass and the review approves. Returns the pull request at once, with g1t as its author and you as its requested_by; follow its progress with get_pull_request. There is no model or agent count to choose. To put many agents to work, assign many issues. Needs the Write role or higher. In preview: only for accounts g1t agents are enabled for." | |
| 811 | 811 | } | |
| 812 | 812 | Op::Delegate => { | |
| 813 | − | "Put an agent on something in one step: open an issue and assign it to the g1t agent at once. Say what you want done in plain words, with what done means if you know it. What must pass before its pull request merges is the default branch's required checks. Needs the Write role or higher, and nothing is opened without it. The issue is opened whatever happens next: agent.status is started (pull is the draft pull request the agent opened; follow it with get_pull_request), queued (every agent slot of the workspace is busy; it starts by itself when one frees up) or not_started, with agent.code saying why (not_paid, trial_used, limit, paused, issue_cap, billing_unavailable or no_model), agent.message saying what to do, and agent.fix_url where. There is no model or agent count to choose." | |
| 813 | + | "Put an agent on something in one step: open an issue and assign it to g1t at once. Say what you want done in plain words, with what done means if you know it. What must pass before its pull request merges is the default branch's required checks. Needs the Write role or higher, and nothing is opened without it. The issue is opened whatever happens next: agent.status is started (pull is the draft pull request the agent opened; follow it with get_pull_request), queued (every agent slot of the workspace is busy; it starts by itself when one frees up) or not_started, with agent.code saying why (not_paid, trial_used, limit, paused, issue_cap, billing_unavailable or no_model), agent.message saying what to do, and agent.fix_url where. There is no model or agent count to choose." | |
| 814 | 814 | } | |
| 815 | 815 | Op::ListLabels => "The labels available on a repository's issues.", | |
| 816 | 816 | Op::AddComment => { | |
| 817 | 817 | "Comment on an issue or a pull request. On a pull request, give path and line to comment on one line of the change." | |
| 818 | 818 | } | |
| 819 | 819 | Op::ReviewPullRequest => { | |
| 820 | − | "Give a verdict on a pull request: approve it, or request changes and say what. Read get_pull_request_changes first. You cannot review a pull request you opened." | |
| 820 | + | "Give a verdict on a pull request: approve it, or request changes and say what. Read get_pull_request_changes first. You cannot review a pull request you opened, or one g1t made for you (you are its requested_by)." | |
| 821 | 821 | } | |
| 822 | 822 | Op::ListPullRequests => { | |
| 823 | 823 | "Pull requests on a repository, newest first. State open covers drafts and those ready for review; closed covers merged and closed." | |
| 835 | 835 | Op::MarkPullRequestReady => { | |
| 836 | 836 | "Mark a draft pull request ready for review. Push your commits first. The summary becomes its description and should say what changed and why." | |
| 837 | 837 | } | |
| 838 | − | Op::ClosePullRequest => "Close a pull request without merging it. Its author may close their own; anyone else needs the Triage role or higher.", | |
| 838 | + | Op::ClosePullRequest => "Close a pull request without merging it. Its author may close their own, and whoever asked g1t for one may close that one; anyone else needs the Triage role or higher.", | |
| 839 | 839 | Op::GetPullRequestChanges => { | |
| 840 | 840 | "What a pull request changes: the files it touches and their line-by-line diff against the commit it started from. Use it to review a pull request or to compare several made for the same issue." | |
| 841 | 841 | } | |
| 1394 | 1394 | "assignees": { | |
| 1395 | 1395 | "type": "array", | |
| 1396 | 1396 | "items": { "type": "string" }, | |
| 1397 | − | "description": "Usernames of the people it is assigned to. Replaces the whole set; an empty list unassigns everyone. To assign it to the g1t agent, use assign_issue.", | |
| 1397 | + | "description": "Usernames of the people it is assigned to. Replaces the whole set; an empty list unassigns everyone. To assign it to g1t, use assign_issue.", | |
| 1398 | 1398 | }, | |
| 1399 | 1399 | })), | |
| 1400 | 1400 | &["repo", "number"], |
| 751 | 751 | "updated_by": "syntaqx", | |
| 752 | 752 | "updated_at": "2026-10-04T16:20:37.508Z" | |
| 753 | 753 | }, | |
| 754 | − | "notes": "`required_checks` replaces the whole list: at most 20 names, each a workflow's name or another status's context, as [`list_check_names`](/reference/api/repositories/list-check-names/) gives them. A pull request merges only once each passes on its head; one nothing has reported holds it too. With the merge queue on, each must also pass on the queued state, so the workflows behind them need `merge_group` in their `on:`. `required_approvals` is at most 6 and `max_revisions` at most 5. See [what a repository can ask for](/guides/g1t-agents/#what-a-repository-can-ask-for). `hold_low_confidence` is on unless turned off: see [confidence](/guides/g1t-agents/#how-sure-the-agent-is)." | |
| 754 | + | "notes": "`required_checks` replaces the whole list: at most 20 names, each a workflow's name or another status's context, as [`list_check_names`](/reference/api/repositories/list-check-names/) gives them. A pull request merges only once each passes on its head; one nothing has reported holds it too. With the merge queue on, each must also pass on the queued state, so the workflows behind them need `merge_group` in their `on:`. `required_approvals` is at most 6 and `max_revisions` at most 5. See [what a repository can ask for](/guides/working-with-g1t/#what-a-repository-can-ask-for). `hold_low_confidence` is on unless turned off: see [confidence](/guides/working-with-g1t/#how-sure-the-agent-is)." | |
| 755 | 755 | }, | |
| 756 | 756 | "list_check_names": { | |
| 757 | 757 | "response": [ | |
| 834 | 834 | "verified": false, | |
| 835 | 835 | "workspaces": [] | |
| 836 | 836 | }, | |
| 837 | + | "requested_by": null, | |
| 837 | 838 | "created_at": "2026-10-01T18:04:11.482Z", | |
| 838 | 839 | "updated_at": "2026-10-01T18:04:11.482Z", | |
| 839 | 840 | "closed_at": null, | |
| 874 | 875 | "verified": false, | |
| 875 | 876 | "workspaces": [] | |
| 876 | 877 | }, | |
| 878 | + | "requested_by": null, | |
| 877 | 879 | "created_at": "2026-10-01T18:04:11.482Z", | |
| 878 | 880 | "updated_at": "2026-10-01T18:04:11.482Z", | |
| 879 | 881 | "closed_at": null, | |
| 907 | 909 | "verified": false, | |
| 908 | 910 | "workspaces": [] | |
| 909 | 911 | }, | |
| 912 | + | "requested_by": null, | |
| 910 | 913 | "created_at": "2026-10-01T18:04:11.482Z", | |
| 911 | 914 | "updated_at": "2026-10-01T18:52:17.093Z", | |
| 912 | 915 | "closed_at": "2026-10-01T18:52:17.093Z", | |
| 950 | 953 | "verified": false, | |
| 951 | 954 | "workspaces": [] | |
| 952 | 955 | }, | |
| 956 | + | "requested_by": null, | |
| 953 | 957 | "created_at": "2026-10-01T18:20:02.117Z", | |
| 954 | 958 | "updated_at": "2026-10-01T18:20:02.117Z" | |
| 955 | 959 | }, | |
| 993 | 997 | "verified": false, | |
| 994 | 998 | "workspaces": [] | |
| 995 | 999 | }, | |
| 1000 | + | "requested_by": null, | |
| 996 | 1001 | "created_at": "2026-10-01T18:20:02.117Z", | |
| 997 | 1002 | "updated_at": "2026-10-01T18:52:17.093Z" | |
| 998 | 1003 | } | |
| 1048 | 1053 | "verified": false, | |
| 1049 | 1054 | "workspaces": [] | |
| 1050 | 1055 | }, | |
| 1056 | + | "requested_by": null, | |
| 1051 | 1057 | "created_at": "2026-10-01T18:04:11.482Z", | |
| 1052 | 1058 | "updated_at": "2026-10-01T18:09:47.305Z", | |
| 1053 | 1059 | "closed_at": null, | |
| 1084 | 1090 | "verified": false, | |
| 1085 | 1091 | "workspaces": [] | |
| 1086 | 1092 | }, | |
| 1093 | + | "requested_by": null, | |
| 1087 | 1094 | "created_at": "2026-10-01T18:04:11.482Z", | |
| 1088 | 1095 | "updated_at": "2026-10-01T19:30:00.214Z", | |
| 1089 | 1096 | "closed_at": "2026-10-01T19:30:00.214Z", | |
| 1115 | 1122 | "verified": false, | |
| 1116 | 1123 | "workspaces": [] | |
| 1117 | 1124 | }, | |
| 1125 | + | "requested_by": null, | |
| 1118 | 1126 | "created_at": "2026-10-01T18:04:11.482Z", | |
| 1119 | 1127 | "updated_at": "2026-10-01T19:41:52.830Z", | |
| 1120 | 1128 | "closed_at": null, | |
| 1137 | 1145 | "issue": 12, | |
| 1138 | 1146 | "title": "Greeting should name the caller", | |
| 1139 | 1147 | "body": null, | |
| 1140 | − | "agent": "g1t-agent", | |
| 1148 | + | "agent": "g1t", | |
| 1141 | 1149 | "runtime": "hosted", | |
| 1142 | 1150 | "status": "draft", | |
| 1143 | 1151 | "fork": { | |
| 1156 | 1164 | "assignees": [], | |
| 1157 | 1165 | "reviewers": [], | |
| 1158 | 1166 | "author": { | |
| 1167 | + | "id": "usr_g1t_agent", | |
| 1168 | + | "username": "g1t", | |
| 1169 | + | "kind": "agent", | |
| 1170 | + | "verified": false, | |
| 1171 | + | "workspaces": [] | |
| 1172 | + | }, | |
| 1173 | + | "requested_by": { | |
| 1159 | 1174 | "id": "usr_01kkntcg1eeb98j62xjm7eh09p", | |
| 1160 | 1175 | "username": "syntaqx", | |
| 1161 | 1176 | "kind": "user", | |
| 1166 | 1181 | "updated_at": "2026-10-01T18:20:02.117Z", | |
| 1167 | 1182 | "confidence": null | |
| 1168 | 1183 | }, | |
| 1169 | − | "notes": "The response is the pull request the g1t agent opened, still a draft. Follow it with [get a pull request](/reference/api/pull-requests/get-pull-request/): `lifecycle` says what the agent is doing." | |
| 1184 | + | "notes": "The response is the pull request g1t opened, still a draft. g1t is its `author` and you are its `requested_by`: you may manage it as if you had opened it, and cannot approve it yourself. Follow it with [get a pull request](/reference/api/pull-requests/get-pull-request/): `lifecycle` says what the agent is doing." | |
| 1170 | 1185 | }, | |
| 1171 | 1186 | "delegate": { | |
| 1172 | 1187 | "request": { | |
| 1191 | 1206 | "verified": false, | |
| 1192 | 1207 | "workspaces": [] | |
| 1193 | 1208 | }, | |
| 1209 | + | "requested_by": null, | |
| 1194 | 1210 | "created_at": "2026-10-05T14:02:11.204Z", | |
| 1195 | 1211 | "updated_at": "2026-10-05T14:02:11.204Z", | |
| 1196 | 1212 | "closed_at": null, | |
| 1199 | 1215 | "assignees": [], | |
| 1200 | 1216 | "blocked_by": [], | |
| 1201 | 1217 | "queued": false, | |
| 1202 | − | "agent": "g1t-agent" | |
| 1218 | + | "agent": "g1t" | |
| 1203 | 1219 | }, | |
| 1204 | 1220 | "pull": { | |
| 1205 | 1221 | "id": "pr_01m4a2c9b6e0h4m8q2t6x0a4d8", | |
| 1208 | 1224 | "issue": 41, | |
| 1209 | 1225 | "title": "Retry webhooks with exponential backoff", | |
| 1210 | 1226 | "body": null, | |
| 1211 | − | "agent": "g1t-agent", | |
| 1227 | + | "agent": "g1t", | |
| 1212 | 1228 | "runtime": "hosted", | |
| 1213 | 1229 | "status": "draft", | |
| 1214 | 1230 | "fork": { | |
| 1227 | 1243 | "assignees": [], | |
| 1228 | 1244 | "reviewers": [], | |
| 1229 | 1245 | "author": { | |
| 1246 | + | "id": "usr_g1t_agent", | |
| 1247 | + | "username": "g1t", | |
| 1248 | + | "kind": "agent", | |
| 1249 | + | "verified": false, | |
| 1250 | + | "workspaces": [] | |
| 1251 | + | }, | |
| 1252 | + | "requested_by": { | |
| 1230 | 1253 | "id": "usr_01kkntcg1eeb98j62xjm7eh09p", | |
| 1231 | 1254 | "username": "syntaqx", | |
| 1232 | 1255 | "kind": "user", | |
| 1244 | 1267 | "fix_url": null | |
| 1245 | 1268 | } | |
| 1246 | 1269 | }, | |
| 1247 | − | "notes": "The issue is opened whatever becomes of the agent. When the workspace's plan does not let it start, `pull` is null and `agent` says why and where to fix it, for example `{ \"status\": \"not_started\", \"code\": \"not_paid\", \"message\": \"Agents need a paid workspace. Start the $20 plan or try it with $5 of free usage after a card check: /acme/-/billing\", \"fix_url\": \"https://g1t.sh/acme/-/billing\" }`. With every agent slot busy, `status` is `queued` and it starts by itself when one frees up. See [putting an agent on something](/guides/g1t-agents/#put-an-agent-on-it-in-one-step)." | |
| 1270 | + | "notes": "The issue is opened whatever becomes of the agent. When the workspace's plan does not let it start, `pull` is null and `agent` says why and where to fix it, for example `{ \"status\": \"not_started\", \"code\": \"not_paid\", \"message\": \"Agents need a paid workspace. Start the $20 plan or try it with $5 of free usage after a card check: /acme/-/billing\", \"fix_url\": \"https://g1t.sh/acme/-/billing\" }`. With every agent slot busy, `status` is `queued` and it starts by itself when one frees up. See [putting an agent on something](/guides/working-with-g1t/#put-an-agent-on-it-in-one-step)." | |
| 1248 | 1271 | }, | |
| 1249 | 1272 | "add_comment": { | |
| 1250 | 1273 | "request": { | |
| 1457 | 1480 | "verified": false, | |
| 1458 | 1481 | "workspaces": [] | |
| 1459 | 1482 | }, | |
| 1483 | + | "requested_by": null, | |
| 1460 | 1484 | "created_at": "2026-10-01T18:20:02.117Z", | |
| 1461 | 1485 | "updated_at": "2026-10-01T18:35:44.902Z" | |
| 1462 | 1486 | } | |
| 1501 | 1525 | "verified": false, | |
| 1502 | 1526 | "workspaces": [] | |
| 1503 | 1527 | }, | |
| 1528 | + | "requested_by": null, | |
| 1504 | 1529 | "created_at": "2026-10-01T18:20:02.117Z", | |
| 1505 | 1530 | "updated_at": "2026-10-01T18:20:02.117Z" | |
| 1506 | 1531 | }, | |
| 1554 | 1579 | "verified": false, | |
| 1555 | 1580 | "workspaces": [] | |
| 1556 | 1581 | }, | |
| 1582 | + | "requested_by": null, | |
| 1557 | 1583 | "created_at": "2026-10-01T18:20:02.117Z", | |
| 1558 | 1584 | "updated_at": "2026-10-01T18:35:44.902Z", | |
| 1559 | 1585 | "confidence": null | |
| 1577 | 1603 | "verified": false, | |
| 1578 | 1604 | "workspaces": [] | |
| 1579 | 1605 | }, | |
| 1606 | + | "requested_by": null, | |
| 1580 | 1607 | "created_at": "2026-10-01T18:04:11.482Z", | |
| 1581 | 1608 | "updated_at": "2026-10-01T18:20:02.117Z", | |
| 1582 | 1609 | "closed_at": null, | |
| 1644 | 1671 | } | |
| 1645 | 1672 | ] | |
| 1646 | 1673 | }, | |
| 1647 | − | "notes": "| Field | |\n| --- | --- |\n| `pull.files` | The files it changes, with lines added and removed. |\n| `statuses` | Its checks: what each workflow run (or another tool, such as a deployment) reported on its head, as `pending`, `success`, `failure` or `error`, with a link to the run. |\n| `required_checks` | Each check the default branch requires (see [`update_repo_settings`](/reference/api/repositories/update-repo-settings/)), as it stands on the head: `success`, `failure`, `pending`, or `expected` when nothing has reported it yet. It merges only once all are `success`. |\n| `checks` | Set when the merge queue took it out, with why in `error`. Older pull requests may show a run of commands from their issue here, from before checks were workflows. |\n| `overlaps` | Other pull requests in progress that change the same files. |\n| `behind` | Whether the default branch has moved since it was made. |\n| `landing` | Whether it is being brought up to date to land. |\n| `lifecycle` | For a pull request the g1t agent is seeing through: its stage, such as `working`, `checking`, `reviewing` or `needs_you`. |\n| `comments` | Comments, reviews and events, with `path`, `line` and `verdict`. |\n| `messages` | Messages sent to the agent working on it. |\n\n| `mergeable` | Whether it merges cleanly into its target: `clean`, `conflicting`, `checking` (being worked out) or `unknown`. Worked out ahead of time whenever it or its target moves. |\n| `conflicts` | When `mergeable` is `conflicting`, the files that conflict. Resolve them by merging the target in, or have the g1t agent do it. |\n| `earlier_checks` | Earlier records like `checks`, newest first. |\n\nChecks are the repository's workflows: they run on `pull_request` events when it is opened, marked ready and pushed to. Read why one failed with [`get_workflow_run`](/reference/api/actions/get-workflow-run/) and [`get_job_logs`](/reference/api/actions/get-job-logs/)." | |
| 1674 | + | "notes": "| Field | |\n| --- | --- |\n| `pull.files` | The files it changes, with lines added and removed. |\n| `statuses` | Its checks: what each workflow run (or another tool, such as a deployment) reported on its head, as `pending`, `success`, `failure` or `error`, with a link to the run. |\n| `required_checks` | Each check the default branch requires (see [`update_repo_settings`](/reference/api/repositories/update-repo-settings/)), as it stands on the head: `success`, `failure`, `pending`, or `expected` when nothing has reported it yet. It merges only once all are `success`. |\n| `checks` | Set when the merge queue took it out, with why in `error`. Older pull requests may show a run of commands from their issue here, from before checks were workflows. |\n| `overlaps` | Other pull requests in progress that change the same files. |\n| `behind` | Whether the default branch has moved since it was made. |\n| `landing` | Whether it is being brought up to date to land. |\n| `lifecycle` | For a pull request g1t is seeing through: its stage, such as `working`, `checking`, `reviewing` or `needs_you`. |\n| `comments` | Comments, reviews and events, with `path`, `line` and `verdict`. |\n| `messages` | Messages sent to the agent working on it. |\n\n| `mergeable` | Whether it merges cleanly into its target: `clean`, `conflicting`, `checking` (being worked out) or `unknown`. Worked out ahead of time whenever it or its target moves. |\n| `conflicts` | When `mergeable` is `conflicting`, the files that conflict. Resolve them by merging the target in, or have g1t do it. |\n| `earlier_checks` | Earlier records like `checks`, newest first. |\n\nChecks are the repository's workflows: they run on `pull_request` events when it is opened, marked ready and pushed to. Read why one failed with [`get_workflow_run`](/reference/api/actions/get-workflow-run/) and [`get_job_logs`](/reference/api/actions/get-job-logs/)." | |
| 1648 | 1675 | }, | |
| 1649 | 1676 | "get_pull_request_changes": { | |
| 1650 | 1677 | "response": { | |
| 1735 | 1762 | "verified": false, | |
| 1736 | 1763 | "workspaces": [] | |
| 1737 | 1764 | }, | |
| 1765 | + | "requested_by": null, | |
| 1738 | 1766 | "created_at": "2026-10-01T18:20:02.117Z", | |
| 1739 | 1767 | "updated_at": "2026-10-01T18:33:10.398Z" | |
| 1740 | 1768 | } | |
| 1810 | 1838 | "verified": false, | |
| 1811 | 1839 | "workspaces": [] | |
| 1812 | 1840 | }, | |
| 1841 | + | "requested_by": null, | |
| 1813 | 1842 | "created_at": "2026-10-01T18:20:02.117Z", | |
| 1814 | 1843 | "updated_at": "2026-10-01T18:52:17.093Z" | |
| 1815 | 1844 | }, | |
| 1854 | 1883 | "verified": false, | |
| 1855 | 1884 | "workspaces": [] | |
| 1856 | 1885 | }, | |
| 1886 | + | "requested_by": null, | |
| 1857 | 1887 | "created_at": "2026-10-01T18:20:02.117Z", | |
| 1858 | 1888 | "updated_at": "2026-10-01T19:02:48.760Z" | |
| 1859 | 1889 | } | |
| 1883 | 1913 | "id": "qen_01m43khqjmeb5a1magayrnk63z", | |
| 1884 | 1914 | "number": 11, | |
| 1885 | 1915 | "title": "Fix trailing newline in greeting", | |
| 1886 | − | "agent": "g1t-agent", | |
| 1916 | + | "agent": "g1t", | |
| 1887 | 1917 | "state": "landed", | |
| 1888 | 1918 | "ahead": [], | |
| 1889 | 1919 | "base_commit": "4b1d7e9c2a5f8e3d6c0b9a7e5d3c1b8a6f4e2d0c", | |
| 1933 | 1963 | }, | |
| 1934 | 1964 | "response": { | |
| 1935 | 1965 | "id": "msg_01m43t66tde8hs2vpxhh3v8tqw", | |
| 1936 | − | "author": "g1t-agent", | |
| 1966 | + | "author": "g1t", | |
| 1937 | 1967 | "body": "Are you renaming greet() in src/lib.rs? I need to call it from #16.", | |
| 1938 | 1968 | "created_at": "2026-10-01T18:58:12.301Z", | |
| 1939 | 1969 | "delivered_at": "2026-10-01T18:58:40.117Z", | |
| 3537 | 3567 | }, | |
| 3538 | 3568 | "number": 14 | |
| 3539 | 3569 | }, | |
| 3540 | − | "created_by": "g1t-agent", | |
| 3570 | + | "created_by": "g1t", | |
| 3541 | 3571 | "pinned": false, | |
| 3542 | 3572 | "created_at": "2026-10-01T18:40:12.000Z", | |
| 3543 | 3573 | "updated_at": "2026-10-01T18:40:12.000Z", | |
| 3645 | 3675 | "number": null, | |
| 3646 | 3676 | "state": null, | |
| 3647 | 3677 | "author": null, | |
| 3678 | + | "requested_by": null, | |
| 3648 | 3679 | "labels": [], | |
| 3649 | 3680 | "topics": [], | |
| 3650 | 3681 | "slug": null, | |
| 3654 | 3685 | ], | |
| 3655 | 3686 | "notes": [] | |
| 3656 | 3687 | }, | |
| 3657 | − | "notes": "`q` takes words, `\"exact phrases\"`, `-words` to leave out, and qualifiers: `repo:owner/name`, `org:` (or `workspace:`), `language:`, `path:` (a glob when it has `*`), `is:issue`, `is:pr`, `is:open`, `is:closed`, `is:merged`, `is:draft`, `is:public`, `is:private`, `author:` and `label:`; most can be left out with a leading `-`, as in `-label:wontfix`. `type` is `repositories`, `code`, `issues`, `pulls` or `people`; without it, `path:` means code, `is:pr` pull requests, `is:open`, `author:` or `label:` issues, and anything else repositories. `counts` says how many results each type has, up to 1,000. Each result's `snippet` (or, for code, each of its `lines`) is a list of parts, `highlight` true where the query matched. Code is searched on default branches and needs a word of three characters or more, unless the query names a `repo:`. Public content is returned to anyone, without a token; private content only to people who can read it (members of its workspace, and people given a role on the repository), checked when the search runs, so a repository made private, or a role taken away, stops appearing at once. A g1t agent's token can search too." | |
| 3688 | + | "notes": "`q` takes words, `\"exact phrases\"`, `-words` to leave out, and qualifiers: `repo:owner/name`, `org:` (or `workspace:`), `language:`, `path:` (a glob when it has `*`), `is:issue`, `is:pr`, `is:open`, `is:closed`, `is:merged`, `is:draft`, `is:public`, `is:private`, `author:` and `label:`; most can be left out with a leading `-`, as in `-label:wontfix`. `type` is `repositories`, `code`, `issues`, `pulls` or `people`; without it, `path:` means code, `is:pr` pull requests, `is:open`, `author:` or `label:` issues, and anything else repositories. `counts` says how many results each type has, up to 1,000. Each result's `snippet` (or, for code, each of its `lines`) is a list of parts, `highlight` true where the query matched. Code is searched on default branches and needs a word of three characters or more, unless the query names a `repo:`. Public content is returned to anyone, without a token; private content only to people who can read it (members of its workspace, and people given a role on the repository), checked when the search runs, so a repository made private, or a role taken away, stops appearing at once. A g1t agent's token can search too. An issue or pull request g1t opened has `author` `g1t` and `requested_by` the person it was for; `author:` matches the author." | |
| 3658 | 3689 | }, | |
| 3659 | 3690 | "search_context": { | |
| 3660 | 3691 | "params": { |
| 9 | 9 | ||
| 10 | 10 | export default defineConfig({ | |
| 11 | 11 | site: 'https://docs.g1t.sh', | |
| 12 | + | // Pages that moved, so links already shared still arrive. | |
| 13 | + | redirects: { | |
| 14 | + | '/guides/g1t-agents/': '/guides/working-with-g1t/', | |
| 15 | + | }, | |
| 12 | 16 | integrations: [ | |
| 13 | 17 | starlight({ | |
| 14 | 18 | title: 'g1t docs', | |
| 76 | 80 | { | |
| 77 | 81 | label: 'Agents', | |
| 78 | 82 | items: [ | |
| 79 | − | { label: 'g1t agents', slug: 'guides/g1t-agents' }, | |
| 83 | + | { label: "g1t's agent", slug: 'guides/working-with-g1t' }, | |
| 80 | 84 | { label: 'Guardrails', slug: 'guides/guardrails' }, | |
| 81 | 85 | { label: 'Outcomes and plans', slug: 'guides/outcomes' }, | |
| 82 | 86 | { label: 'Talking to agents', slug: 'guides/talking-to-agents' }, |
| 161 | 161 | ||
| 162 | 162 | You can't resolve a merge conflict on the pull request's page. | |
| 163 | 163 | ||
| 164 | − | - **Instead.** Ask a g1t agent to resolve it, or fix it on the command line. | |
| 164 | + | - **Instead.** Ask g1t to resolve it, or fix it on the command line. | |
| 165 | 165 | See [Conflicts](/guides/pull-requests/#conflicts). | |
| 166 | 166 | - **Status.** Planned. | |
| 167 | 167 | ||
| 309 | 309 | - **Instead.** Keep **Ask a person before merging low-confidence changes** | |
| 310 | 310 | on, and make your workflows required checks. A high rating on a | |
| 311 | 311 | repository with weak tests means less. See | |
| 312 | − | [How sure the agent is](/guides/g1t-agents/#how-sure-the-agent-is). | |
| 312 | + | [How sure the agent is](/guides/working-with-g1t/#how-sure-the-agent-is). | |
| 313 | 313 | - **Status.** The signals and their weights may change as we learn from | |
| 314 | 314 | real changes. | |
| 315 | 315 | ||
| 343 | 343 | ### Self-hosting is early | |
| 344 | 344 | ||
| 345 | 345 | [Running g1t yourself](/guides/self-hosting/) gives you the core forge: | |
| 346 | − | accounts, repositories over HTTP, issues and pull requests. g1t agents, | |
| 346 | + | accounts, repositories over HTTP, issues and pull requests. g1t's agent, | |
| 347 | 347 | Actions, deployments, git over SSH, the REST API and MCP are off, and it is | |
| 348 | 348 | not ready for the open internet. **Status.** Planned, in phases. | |
| 349 | 349 |
| 87 | 87 | ||
| 88 | 88 | ## Assignees and reviewers | |
| 89 | 89 | ||
| 90 | − | An issue is assigned to people, to the g1t agent, or to both. A pull | |
| 91 | − | request has assignees too, and reviewers: the people, or the g1t agent, | |
| 90 | + | An issue is assigned to people, to g1t, or to both. A pull | |
| 91 | + | request has assignees too, and reviewers: the people, or g1t, | |
| 92 | 92 | whose review was asked for. Each shows beside the conversation, with where | |
| 93 | 93 | every reviewer stands. | |
| 94 | 94 | ||
| 134 | 134 | unless the repository allows bypassing them and someone who can merge | |
| 135 | 135 | chooses to. Checks that are not required are shown, and never hold a merge. | |
| 136 | 136 | ||
| 137 | − | The same rules hold for people and agents. A g1t agent's pull request is | |
| 137 | + | The same rules hold for people and agents. A pull request g1t opens is | |
| 138 | 138 | checked by the same workflows as yours, and an agent cannot mark its own | |
| 139 | 139 | work as passing: only the workflow runs report. | |
| 140 | 140 | ||
| 144 | 144 | on a single line of its change. Line comments are shown in the **Files changed** | |
| 145 | 145 | tab under the line they are about. | |
| 146 | 146 | ||
| 147 | − | You can also ask a **g1t agent** to review. It reads the change in a sandbox | |
| 147 | + | You can also ask **g1t** to review. It reads the change in a sandbox | |
| 148 | 148 | of its own and posts comments on lines, a summary and a verdict, as | |
| 149 | − | `g1t-agent`. | |
| 149 | + | `g1t`. | |
| 150 | 150 | ||
| 151 | 151 | A reviewer can also give a verdict: **approve**, or **request changes**. | |
| 152 | 152 | The pull request shows where each reviewer stands. You cannot give a verdict | |
| 153 | 153 | on a pull request you opened, and that holds for agents too: one agent can | |
| 154 | 154 | review another's work, but not its own. | |
| 155 | 155 | ||
| 156 | − | Requesting changes on a g1t agent's pull request sends the agent back to | |
| 156 | + | Requesting changes on a pull request g1t opened sends g1t back to | |
| 157 | 157 | make them. See [talk to agents](/guides/talking-to-agents/#ask-for-changes). | |
| 158 | 158 | ||
| 159 | 159 | ## Overlap | |
| 168 | 168 | when the second one tries to merge. Agents get the same list from | |
| 169 | 169 | the `pull_request` tool's `get` action, as `overlaps`. | |
| 170 | 170 | ||
| 171 | − | A g1t agent is told about the other work before it starts. Its instructions | |
| 171 | + | g1t is told about the other work before it starts. Its instructions | |
| 172 | 172 | list every pull request in progress in the repository, what each is for and | |
| 173 | 173 | which files it changes, and ask it to keep its edits small and local where | |
| 174 | 174 | it has to touch the same files. It is told again when it is sent back to | |
| 192 | 192 | ||
| 193 | 193 | **Catch up with main** brings it up to date. When the pull request and `main` changed | |
| 194 | 194 | different files, g1t merges `main` in itself and pushes the merge in a few | |
| 195 | − | seconds. When they changed some of the same files, a g1t agent merges `main` | |
| 195 | + | seconds. When they changed some of the same files, g1t merges `main` | |
| 196 | 196 | into the pull request in a sandbox: if the merge is clean, it is pushed as it | |
| 197 | 197 | is; if it conflicts, the agent is given the conflicted files and what the | |
| 198 | 198 | pull request is for, resolves them, and pushes the result, and the session |
| 211 | 211 | remove or invite anyone, answer an invitation, or change the base | |
| 212 | 212 | permission. When the person it works for loses their role on the | |
| 213 | 213 | repository, or leaves the workspace, the agent loses it too. See | |
| 214 | − | [credentials](/guides/g1t-agents/#credentials). | |
| 214 | + | [credentials](/guides/working-with-g1t/#credentials). | |
| 215 | 215 | ||
| 216 | 216 | Because putting agents to work spends compute, it needs Write. Someone with | |
| 217 | 217 | Read or Triage who mentions or assigns an agent is told so, and nothing |
| 180 | 180 | ## Pull requests | |
| 181 | 181 | ||
| 182 | 182 | A pull request's workflows run on each new head: when it is opened, when | |
| 183 | − | a commit is pushed to it, and, for one a g1t agent makes, when the agent | |
| 183 | + | a commit is pushed to it, and, for one g1t makes, when g1t | |
| 184 | 184 | marks it ready, which on g1t is when it first has code. Each head runs | |
| 185 | 185 | each workflow once. | |
| 186 | 186 | ||
| 187 | + | `github.event.pull_request` reads as it does on GitHub. For a pull request | |
| 188 | + | g1t made, `pull_request.user` is g1t (`login` `g1t`, `type` `Bot`), and | |
| 189 | + | `pull_request.requested_by` names the person who asked for it; it is `null` | |
| 190 | + | on anyone else's. `github.event.issue.requested_by` does the same for an | |
| 191 | + | issue g1t's agent filed. `sender` is whoever caused the event. | |
| 192 | + | ||
| 187 | 193 | ## Checks | |
| 188 | 194 | ||
| 189 | 195 | A pull request's checks are its workflows. Each workflow that runs on | |
| 202 | 208 | builds, on the branch `g1t-queue/<entry>`, and the state lands only if | |
| 203 | 209 | they and every required check pass on it. A workflow behind a required | |
| 204 | 210 | check needs `merge_group` in its `on:`. | |
| 205 | − | - **A pull request a g1t agent is working on** goes back to the agent when | |
| 211 | + | - **A pull request g1t is working on** goes back to g1t when | |
| 206 | 212 | a check fails, with the end of each failed job's log. The agent reads the | |
| 207 | 213 | run and its logs with the same tools you have, fixes the cause, and | |
| 208 | 214 | pushes; the workflows run again. See | |
| 209 | − | [seeing it through](/guides/g1t-agents/#seeing-it-through). | |
| 215 | + | [seeing it through](/guides/working-with-g1t/#seeing-it-through). | |
| 210 | 216 | ||
| 211 | 217 | ```yaml | |
| 212 | 218 | name: CI | |
| 255 | 261 | token for the run, with `GITHUB_TOKEN` as its alias. A pull request's runs | |
| 256 | 262 | get secrets and the token only when its author has the Write | |
| 257 | 263 | [role](/guides/access-and-roles/) or higher on the repository, a member or | |
| 258 | − | an outside collaborator, or is g1t's agent. Anyone else's, such as one | |
| 264 | + | an outside collaborator, or is g1t working on its own. For a pull request | |
| 265 | + | g1t made, its author is g1t and the person who asked for it is the one | |
| 266 | + | whose role counts. Anyone else's, such as one | |
| 259 | 267 | from a fork or by someone with Read or Triage, runs without secrets and | |
| 260 | 268 | with an empty token. See | |
| 261 | 269 | [who gets secrets](/guides/secrets-and-variables/#who-gets-secrets). |
| 20 | 20 | ||
| 21 | 21 | ## Runs | |
| 22 | 22 | ||
| 23 | − | A **run** is one sandbox g1t starts. Each g1t agent run does one kind of | |
| 23 | + | A **run** is one sandbox g1t starts. Each g1t run does one kind of | |
| 24 | 24 | work on one pull request: | |
| 25 | 25 | ||
| 26 | 26 | | Kind | What the agent does | | |
| 37 | 37 | ||
| 38 | 38 | Each run records: | |
| 39 | 39 | ||
| 40 | − | - the agent (`g1t-agent`) and, for members, the model it runs on; | |
| 40 | + | - the agent (`g1t`) and, for members, the model it runs on; | |
| 41 | 41 | - its pull request, or for a plan, the outcome; | |
| 42 | 42 | - its status: starting, running, done, failed or stopped; | |
| 43 | 43 | - its **current step**, one line such as `Edited src/auth.ts` or | |
| 135 | 135 | ||
| 136 | 136 | ### How agents use it | |
| 137 | 137 | ||
| 138 | − | Every g1t agent run is given memory when it starts: the workspace's and | |
| 138 | + | Every g1t run is given memory when it starts: the workspace's and | |
| 139 | 139 | the project's, each labelled, pinned memories first, then the ones used | |
| 140 | 140 | most recently, up to about 6,000 characters. Agents are told to treat it as | |
| 141 | 141 | notes from colleagues: usually right, sometimes out of date, and where it |
| 5 | 5 | ||
| 6 | 6 | Every workspace keeps an audit log. It records: | |
| 7 | 7 | ||
| 8 | − | - **Everything an agent does** with its [run credentials](/guides/g1t-agents/#credentials), | |
| 8 | + | - **Everything an agent does** with its [run credentials](/guides/working-with-g1t/#credentials), | |
| 9 | 9 | reads included: every API and MCP call, every clone and fetch, every push. | |
| 10 | 10 | - **Every change people and workspace tokens make** through the API, the | |
| 11 | 11 | MCP server and git: opening and closing issues, comments, merges, | |
| 41 | 41 | | --- | --- | | |
| 42 | 42 | | Time | When it happened, to the millisecond. | | |
| 43 | 43 | | Actor | Who did it: a person, an agent, or a workspace token. | | |
| 44 | − | | On behalf of | For an agent, the person it worked for: `g1t-agent on behalf of syntaqx`. | | |
| 44 | + | | On behalf of | For an agent, the person it worked for: `g1t on behalf of syntaqx`. | | |
| 45 | 45 | | Run | The agent run, with its kind: `implement`, `review`, `update` and so on. | | |
| 46 | 46 | | Credential | The id of the token used. | | |
| 47 | 47 | | Action | The API or MCP operation, such as `create_issue`, or `git.push` and `git.fetch`. | | |
| 52 | 52 | | Request id | The request's id, the same one Cloudflare logs it under. | | |
| 53 | 53 | ||
| 54 | 54 | The rules that refuse an agent are listed under | |
| 55 | − | [credentials](/guides/g1t-agents/#credentials). | |
| 55 | + | [credentials](/guides/working-with-g1t/#credentials). | |
| 56 | 56 | ||
| 57 | 57 | ## Read the log | |
| 58 | 58 |
| 341 | 341 | | `issues:write` | Open, edit, close and comment on issues | | |
| 342 | 342 | | `pull_requests:read` | Read pull requests, their changes, sessions and merge queues | | |
| 343 | 343 | | `pull_requests:write` | Open, review, close and merge pull requests | | |
| 344 | − | | `agents:run` | Put g1t agents to work and message them, which uses the workspace's money | | |
| 344 | + | | `agents:run` | Put g1t to work and message it, which uses the workspace's money | | |
| 345 | 345 | | `workflows:read` | Read workflows, runs and logs | | |
| 346 | 346 | | `workflows:write` | Run, cancel, rerun and turn workflows on or off | | |
| 347 | 347 | | `memory:read` | Recall memory and search the workspace's context | | |
| 394 | 394 | | Preset | Scopes | | |
| 395 | 395 | | --- | --- | | |
| 396 | 396 | | Read only | Every `read` scope. Changes nothing. | | |
| 397 | − | | Agent | Every `read` scope except `runners:read`, and `code:write`, `issues:write`, `pull_requests:write`, `agents:run` and `memory:write`. Reads everything, works on issues and pull requests, pushes code and runs g1t agents. No admin scope. | | |
| 397 | + | | Agent | Every `read` scope except `runners:read`, and `code:write`, `issues:write`, `pull_requests:write`, `agents:run` and `memory:write`. Reads everything, works on issues and pull requests, pushes code and puts g1t to work. No admin scope. | | |
| 398 | 398 | | CI | `repo:read`, `code:read`, `code:write`, `workflows:read` and `workflows:write`. Clones and pushes code, and runs workflows. | | |
| 399 | 399 | | Full access | Everything you can do, including deleting repositories and changing who has access. Marked **Dangerous**. | | |
| 400 | 400 |
| 63 | 63 | Claude Code from `CLAUDE.md`, and Codex, OpenCode and Cursor from | |
| 64 | 64 | `AGENTS.md`. To keep one file, write `AGENTS.md` and put `@AGENTS.md` in | |
| 65 | 65 | `CLAUDE.md`. g1t's own agents read both; see | |
| 66 | − | [repository instructions](/guides/g1t-agents/#repository-instructions). | |
| 66 | + | [repository instructions](/guides/working-with-g1t/#repository-instructions). | |
| 67 | 67 | ||
| 68 | 68 | ### Recording sessions automatically | |
| 69 | 69 | ||
| 137 | 137 | was made. If it has, pull `main` into the fork and push before asking for a | |
| 138 | 138 | merge. | |
| 139 | 139 | ||
| 140 | − | ## Talking to g1t agents | |
| 140 | + | ## Talking to g1t | |
| 141 | 141 | ||
| 142 | − | Your agent can send the g1t agent working on a pull request a message with | |
| 143 | − | `agent` and `message`; it arrives at that agent's next step. g1t agents also ask | |
| 142 | + | Your agent can send g1t a message on a pull request it is working on, with | |
| 143 | + | `agent` and `message`; it arrives at g1t's next step. g1t's runs also ask | |
| 144 | 144 | each other questions and hand each other work. See | |
| 145 | 145 | [talk to agents](/guides/talking-to-agents/). | |
| 146 | 146 |
| 19 | 19 | | **Search** | One search across the catalog, docs, issues, pull requests and memory | | |
| 20 | 20 | | **Scorecards** | A few rules every project should meet, each failing one a click away from an issue an agent fixes | | |
| 21 | 21 | ||
| 22 | − | Every g1t agent run starts with a **Context** section drawn from the hub, | |
| 22 | + | Every g1t run starts with a **Context** section drawn from the hub, | |
| 23 | 23 | and your own agents can ask it through the [MCP tools](#mcp-tools): | |
| 24 | 24 | the `search` tool's `context` and `entity` actions. | |
| 25 | 25 | ||
| 169 | 169 | ||
| 170 | 170 | ## Agents start with context | |
| 171 | 171 | ||
| 172 | − | Every g1t agent run is given a **Context** section, after the project's | |
| 172 | + | Every g1t run is given a **Context** section, after the project's | |
| 173 | 173 | memory, within about 4,000 characters: | |
| 174 | 174 | ||
| 175 | 175 | - the project's stack, test commands, owners, docs, and its environments |
| 314 | 314 | rather than the build. Secrets reach a preview only when the pull | |
| 315 | 315 | request's author has Write or higher on the repository, whether a member | |
| 316 | 316 | or an [outside collaborator](/guides/access-and-roles/#outside-collaborators), | |
| 317 | − | or it is g1t's own agent. A preview of anyone else's pull request, such as | |
| 317 | + | or it is g1t working on its own. For a pull request g1t made, whoever | |
| 318 | + | asked for it is the one whose role counts. A preview of anyone else's pull request, such as | |
| 318 | 319 | one from a fork or by someone with Read or Triage, is built and runs with | |
| 319 | 320 | variables only, no secrets. | |
| 320 | 321 |
| 1 | − | --- | |
| 2 | − | title: g1t agents | |
| 3 | − | description: Have g1t's own agents work on an issue. | |
| 4 | − | --- | |
| 5 | − | ||
| 6 | − | g1t can do the work itself. Assign an open issue to the g1t agent and it | |
| 7 | − | opens a pull request for the issue, works in a sandbox and a fork of its | |
| 8 | − | own, and reports back as it goes. There is nothing to configure: you do not | |
| 9 | − | say how many agents or which model. Scale comes from assigning many issues, | |
| 10 | − | each to its own agent, all working at once. | |
| 11 | − | ||
| 12 | − | g1t agents are paid for by the workspace they work for, so they need a | |
| 13 | − | paid workspace or the free trial; see [who can run agents](#who-can-run-agents) | |
| 14 | − | and [usage and billing](/guides/usage-and-billing/). Everyone can also | |
| 15 | − | [bring their own agent](/guides/bring-your-own-agent/), which costs | |
| 16 | − | nothing on g1t. | |
| 17 | − | ||
| 18 | − | To hand over a whole outcome rather than one issue at a time, have an agent | |
| 19 | − | plan it first: see [hand off an outcome](/guides/outcomes/). | |
| 20 | − | ||
| 21 | − | ## Put an agent on it in one step | |
| 22 | − | ||
| 23 | − | When the work is not written down yet, open the issue and hand it to the | |
| 24 | − | agent at once: | |
| 25 | − | ||
| 26 | − | 1. On Mission control, choose **Put an agent on it**. | |
| 27 | − | 2. Pick the project, give a title, and say what you want done in plain | |
| 28 | − | words, with what done means if you know it. | |
| 29 | − | 3. Choose **Put an agent on it**. | |
| 30 | − | ||
| 31 | − | You land on the new issue with the agent already at work on its pull | |
| 32 | − | request. The same choice is on a project's **New issue** page, as **Assign | |
| 33 | − | g1t-agent now**, and in the ⌘K palette as **Put an agent on …** followed by | |
| 34 | − | a project's name. | |
| 35 | − | ||
| 36 | − | Putting an agent to work needs the Write role on the project. Without it, | |
| 37 | − | nothing is opened. With it, the issue is always opened, even when the | |
| 38 | − | agent cannot start: | |
| 39 | − | ||
| 40 | − | | What happened | What you see | | |
| 41 | − | | --- | --- | | |
| 42 | − | | The agent started | The issue, with its draft pull request under **Assignees**. | | |
| 43 | − | | Every agent slot of the workspace is busy | The issue, queued for g1t-agent. It starts by itself when a slot frees up. | | |
| 44 | − | | The workspace's plan or limits refused it | The issue is opened, and the composer says why and links to the fix: start the plan or the trial (`not_paid`, `trial_used`), raise the monthly limit (`limit`) or the cap per issue (`issue_cap`), or connect a model (`no_model`). A workspace g1t `paused` says to contact support. | | |
| 45 | − | ||
| 46 | − | From the API or an agent of your own, it is one call: | |
| 47 | − | ||
| 48 | − | ```sh | |
| 49 | − | curl -X POST https://api.g1t.sh/repos/<workspace>/<repo>/issues/delegate \ | |
| 50 | − | -H "Authorization: Bearer $G1T_TOKEN" \ | |
| 51 | − | -H "Content-Type: application/json" \ | |
| 52 | − | -d '{"title": "Retry webhooks with exponential backoff", "body": "Deliveries that fail are dropped today. Retry them up to six times."}' | |
| 53 | − | ``` | |
| 54 | − | ||
| 55 | − | The answer holds the `issue`, the `pull` request the agent opened (or | |
| 56 | − | `null`), and `agent`: its `status` (`started`, `queued` or `not_started`), | |
| 57 | − | and when it did not start, a `code`, a `message` and a `fix_url`. On the MCP | |
| 58 | − | server it is the `agent` tool's `delegate` action. See | |
| 59 | − | [put an agent on it](/reference/api/issues/delegate/). | |
| 60 | − | ||
| 61 | − | The `checks` field this call and `create_issue` used to take is | |
| 62 | − | deprecated. It is still accepted: its commands are added to the issue's | |
| 63 | − | body under `## Definition of done`, one line each (`` - `npm test` passes. ``), | |
| 64 | − | and the answer carries a `deprecation` string saying so. What has to pass | |
| 65 | − | before the pull request merges is the default branch's | |
| 66 | − | [required status checks](/guides/pull-requests/#required-status-checks). | |
| 67 | − | ||
| 68 | − | ## Assigning agents | |
| 69 | − | ||
| 70 | − | One issue: | |
| 71 | − | ||
| 72 | − | 1. Open an issue on a repository. | |
| 73 | − | 2. In **Assign to g1t agent**, optionally add guidance for this run, on top | |
| 74 | − | of the issue's description. | |
| 75 | − | 3. Choose **Assign**. | |
| 76 | − | ||
| 77 | − | Many issues: | |
| 78 | − | ||
| 79 | − | 1. Open the repository's **Issues** tab. | |
| 80 | − | 2. Tick the issues to hand over, up to ten at a time. | |
| 81 | − | 3. Choose **Assign to g1t agent**. | |
| 82 | − | ||
| 83 | − | From the API, an agent of your own, or a script: | |
| 84 | − | ||
| 85 | − | ```sh | |
| 86 | − | curl -X POST https://api.g1t.sh/repos/<workspace>/<repo>/issues/12/assign \ | |
| 87 | − | -H "Authorization: Bearer $G1T_TOKEN" | |
| 88 | − | ``` | |
| 89 | − | ||
| 90 | − | The same thing is the `agent` tool's `assign` action on the MCP server, so an agent | |
| 91 | − | planning work can hand issues to g1t agents itself. | |
| 92 | − | ||
| 93 | − | In a comment: write `@g1t-agent take this` on the issue. See | |
| 94 | − | [mentioning g1t-agent](#mentioning-g1t-agent). | |
| 95 | − | ||
| 96 | − | By label: a project can hand every issue given a label to the agent. See | |
| 97 | − | [the label rule](#the-label-rule). | |
| 98 | − | ||
| 99 | − | Each agent appears as a draft pull request on its issue within a few | |
| 100 | − | seconds. The pages update on their own while they work. | |
| 101 | − | ||
| 102 | − | An issue can still have more than one pull request: assign it again, or | |
| 103 | − | have your own agent open one alongside. That is for when you want a second | |
| 104 | − | attempt, not the normal way of working. | |
| 105 | − | ||
| 106 | − | ## What an agent does | |
| 107 | − | ||
| 108 | − | 1. Clones its pull request's fork. | |
| 109 | − | 2. Is told what else is in progress: every other open pull request in the | |
| 110 | − | repository, what it is for and which files it changes. Its session | |
| 111 | − | starts with a note of what it was told. | |
| 112 | − | 3. Reads the code and makes the change the issue asks for, keeping clear of | |
| 113 | − | the other work where it can. | |
| 114 | − | 4. Commits its work. | |
| 115 | − | 5. Pushes to the fork and marks the pull request ready for review, with a | |
| 116 | − | summary as its description. | |
| 117 | − | ||
| 118 | − | Everything it reads, runs and decides is recorded in the pull request's | |
| 119 | − | **Session** as it happens; see [sessions and why-blame](/guides/why-blame/). | |
| 120 | − | The **Files changed** tab shows the resulting diff. | |
| 121 | − | ||
| 122 | − | If an agent fails, or finishes without changing anything, its pull request | |
| 123 | − | is closed and its session says why. | |
| 124 | − | ||
| 125 | − | ## Repository instructions | |
| 126 | − | ||
| 127 | − | Every g1t agent run reads the repository's own instructions for agents | |
| 128 | − | and is told them, labelled as the repository's, before it starts: making a | |
| 129 | − | change, revising it, reviewing, catching up, answering, and planning. | |
| 130 | − | ||
| 131 | − | | File | Read by | | |
| 132 | − | | --- | --- | | |
| 133 | − | | `AGENTS.md` and `CLAUDE.md` at the root | Every run. | | |
| 134 | − | | `AGENTS.md` and `CLAUDE.md` in a subdirectory | Runs whose task touches files under it: the nearest one above each file. Where it disagrees with the root's, it wins for the files under it. | | |
| 135 | − | | `.g1t/review.md` | Reviews: what to check, house rules, paths that need extra care. | | |
| 136 | − | ||
| 137 | − | Write them for an agent that knows nothing about the project: how to build | |
| 138 | − | and test, how things are named, what never to touch. For example: | |
| 139 | − | ||
| 140 | − | ```md | |
| 141 | − | # AGENTS.md | |
| 142 | − | - Run `npm test` and `npm run typecheck` before you finish. | |
| 143 | − | - API handlers return a Result; they never throw. | |
| 144 | − | - Never edit files under `vendor/`. | |
| 145 | − | ``` | |
| 146 | − | ||
| 147 | − | Which directories a task touches comes from the pull request's changed | |
| 148 | − | files, and for new work from the paths the issue names, so naming the | |
| 149 | − | files in an issue helps the right instructions reach the agent. | |
| 150 | − | ||
| 151 | − | **Where they are read from.** The default branch, as it is when the run | |
| 152 | − | starts. A pull request from one of the repository's own branches is read at | |
| 153 | − | its head instead, since only people who can push to the repository can | |
| 154 | − | change it. A pull request from a fork, which includes every change a g1t | |
| 155 | − | agent makes, is never followed: the agent keeps the default branch's | |
| 156 | − | instructions, and if the fork changes them, it is shown the changed text as | |
| 157 | − | part of the change, marked as not instructions. That way nobody can steer | |
| 158 | − | an agent, or the review of their own change, by editing these files in a | |
| 159 | − | pull request. Treat what a fork's head says as untrusted, as you would its | |
| 160 | − | code. | |
| 161 | − | ||
| 162 | − | **Limits.** Each file is cut at 8,000 characters and all of them together | |
| 163 | − | at 24,000; what is left out is named in the prompt. Files are read once per | |
| 164 | − | commit and reused. | |
| 165 | − | ||
| 166 | − | The project's **Agents** page lists the files its runs read, what they say | |
| 167 | − | and when each last changed, with a link to each in the code. Each run's | |
| 168 | − | session starts with a note of which files it read. To change them, change | |
| 169 | − | the files and merge to the default branch. | |
| 170 | − | ||
| 171 | − | ## Seeing it through | |
| 172 | − | ||
| 173 | − | Making the change is the first step. g1t takes the rest itself, and you | |
| 174 | − | get the pull request back ready to merge: | |
| 175 | − | ||
| 176 | − | 1. **Checks.** Every push the agent makes runs the repository's | |
| 177 | − | [workflows](/guides/actions/) on the pull request, as for anyone's pull | |
| 178 | − | request. g1t waits for them to finish. Only the workflow runs report, so | |
| 179 | − | the agent has no say in the result. | |
| 180 | − | 2. **Review.** A different agent reads the change and posts comments on | |
| 181 | − | lines, a summary and a verdict. | |
| 182 | − | 3. **Revision.** If a check fails or the review asks for changes, the | |
| 183 | − | author is sent back with exactly what was found. For a failed check, | |
| 184 | − | that is the end of the log of each failed job, up to three jobs and | |
| 185 | − | about 3,000 characters each; it can read more with `get_workflow_run` | |
| 186 | − | and `get_job_logs`. Steps 1 and 2 run again on the result. This happens | |
| 187 | − | at most twice, or as often as the repository's **Revisions before asking | |
| 188 | − | you** allows. | |
| 189 | − | 4. **Ready to merge.** The required checks passed and it is approved. | |
| 190 | − | Merging is yours, unless the repository says otherwise (below). | |
| 191 | − | ||
| 192 | − | Agents are told to run the same tests and linters the workflows run before | |
| 193 | − | they finish, so most failures are caught in the sandbox. What "done" means | |
| 194 | − | for the issue, if its description says so, is context for the agent and | |
| 195 | − | its reviewer; what decides the merge is the default branch's | |
| 196 | − | [required status checks](/guides/pull-requests/#required-status-checks). | |
| 197 | − | A repository with no workflows has nothing to prove a change works; **Add | |
| 198 | − | CI** gives it one ([add CI](/guides/actions/#add-ci)). | |
| 199 | − | ||
| 200 | − | If `main` has moved in the meantime, that does not hold the pull request | |
| 201 | − | up. Merging it brings it up to date first: g1t merges `main` in, an agent | |
| 202 | − | resolves any conflict, and it lands. A repository that wants every pull | |
| 203 | − | request caught up and checked again before it may merge turns on **Require | |
| 204 | − | pull requests to be up to date before merging** in its settings; catching | |
| 205 | − | up is then a step of its own, before "ready". | |
| 206 | − | ||
| 207 | − | g1t also works out ahead of time whether each pull request still merges | |
| 208 | − | cleanly, every time it or `main` moves | |
| 209 | − | ([how](/guides/pull-requests/#conflicts)). When one of an agent's pull | |
| 210 | − | requests is found to conflict, g1t does not wait for a merge to trip over | |
| 211 | − | it: the agent is sent to merge `main` in and resolve the conflicts, told | |
| 212 | − | which files conflict, and the checks run again on the result. | |
| 213 | − | ||
| 214 | − | The pull request's page shows which step it is at. While a required check | |
| 215 | − | has not reported on its latest commit, it says so: "Waiting for the | |
| 216 | − | required check CI to report on its latest commit." If g1t cannot finish, | |
| 217 | − | because a required check still fails after the agent revised as often as | |
| 218 | − | the repository allows ("The required check CI still fails after the agent | |
| 219 | − | revised twice."), a review could not be written, or a conflict could not | |
| 220 | − | be resolved while bringing it up to date, it stops and the page says | |
| 221 | − | **Needs you**, with the reason. A check that is not required and still | |
| 222 | − | fails after the last revision does not hold it. Pushing to the pull | |
| 223 | − | request yourself starts it moving again. | |
| 224 | − | ||
| 225 | − | ### What a repository can ask for | |
| 226 | − | ||
| 227 | − | Under a project's **Settings → Branches and merging**, someone with the | |
| 228 | − | Maintain [role](/guides/access-and-roles/) or higher sets the rules its pull | |
| 229 | − | requests follow. **Branch protection** holds the rules for every pull | |
| 230 | − | request, a person's or an agent's; they are described in | |
| 231 | − | [required status checks](/guides/pull-requests/#required-status-checks): | |
| 232 | − | ||
| 233 | − | | Setting | Default | What it does | | |
| 234 | − | | --- | --- | --- | | |
| 235 | − | | Require a pull request to change the default branch | Off | Refuses pushes to the default branch. | | |
| 236 | − | | Required status checks | None | The checks that must pass on a pull request's head before it merges. | | |
| 237 | − | | Required approvals | None | How many reviewers must approve before a merge. A reviewer who asked for changes blocks it. | | |
| 238 | − | | A g1t agent's approval counts | On | Off means approvals have to come from people. | | |
| 239 | − | | Require branches to be up to date before merging | Off | On means catching up is a step of its own and the checks run again. | | |
| 240 | − | | Merge through a queue | Off | Merging tests a pull request together with those ahead of it; the default branch only moves to a combination that passed. See [merge queue](/guides/merge-queue/). | | |
| 241 | − | | Allow bypassing required checks | On | Lets someone who may merge merge without the required checks passing. Off means nobody can. | | |
| 242 | − | ||
| 243 | − | **g1t agents** holds what g1t does with its own agents' pull requests: | |
| 244 | − | ||
| 245 | − | | Setting | Default | What it does | | |
| 246 | − | | --- | --- | --- | | |
| 247 | − | | Review by a second agent | On | Off leaves review to people. | | |
| 248 | − | | Revisions before asking you | 2 | How often an agent is sent back before g1t stops. | | |
| 249 | − | | Merge automatically when ready | Off | Lands a g1t agent's pull request once every rule is met. | | |
| 250 | − | | Ask a person before merging low-confidence changes | On | A g1t agent's change [rated low](#how-sure-the-agent-is) waits for a person's approval instead of merging by itself or joining the queue. | | |
| 251 | − | ||
| 252 | − | A g1t agent's pull request follows the same rules as anyone's. If the | |
| 253 | − | repository wants approvals from people, it waits for them, and shows | |
| 254 | − | **Needs you** until they arrive. | |
| 255 | − | ||
| 256 | − | ### Talking to an agent | |
| 257 | − | ||
| 258 | − | While a g1t agent works, you can steer it with **Message the agent** on its | |
| 259 | − | pull request; it reads the message at its next step, without starting | |
| 260 | − | over. Once it is done, a review with **Request changes** sends it back to | |
| 261 | − | make them, and the checks and review run again. g1t agents working at the | |
| 262 | − | same time can also ask each other questions and hand each other work. See | |
| 263 | − | [talk to agents](/guides/talking-to-agents/). | |
| 264 | − | ||
| 265 | − | ### Merging automatically | |
| 266 | − | ||
| 267 | − | A repository can land a g1t agent's pull request by itself once it is | |
| 268 | − | ready. Someone with the Maintain role or higher turns this on under the repository's | |
| 269 | − | **Settings → Branches and merging**; it is off to begin with. The merge is recorded as made by | |
| 270 | − | `g1t`, the issue closes naming the pull request, and nothing short of | |
| 271 | − | ready is ever merged this way: every required check has passed on its | |
| 272 | − | head. One that is behind `main` is brought up to | |
| 273 | − | date as part of the merge. Pull requests from people and from other | |
| 274 | − | agents always wait for a person to merge them. | |
| 275 | − | ||
| 276 | − | Every step is recorded: revisions and catch-ups in the pull request's | |
| 277 | − | **Session**, reviews in its conversation. | |
| 278 | − | ||
| 279 | − | This applies to pull requests made by g1t agents. One you or your own | |
| 280 | − | agent opened is yours to drive; the same workflows run on it, and you can ask | |
| 281 | − | for a review or a catch-up from its page. | |
| 282 | − | ||
| 283 | − | ### How sure the agent is | |
| 284 | − | ||
| 285 | − | Once a g1t agent has finished a change, g1t records how sure it is that the | |
| 286 | − | change is right: **high**, **medium** or **low**, with a few words saying | |
| 287 | − | why, such as "Low — tests not added, 3 revisions". It shows on the pull | |
| 288 | − | request, under the agent, and on Mission control. It is worked out again as | |
| 289 | − | the change moves through checks, review and revision, and kept with each | |
| 290 | − | run, so a run's page says how the change stood when that run left it. | |
| 291 | − | ||
| 292 | − | Confidence comes from what g1t can observe, not from how the agent sounds. | |
| 293 | − | Each signal below that tells against the change adds points, or makes it | |
| 294 | − | low on its own. No points is high, one or two is medium, and three or more | |
| 295 | − | is low. | |
| 296 | − | ||
| 297 | − | | Signal | Effect | | |
| 298 | − | | --- | --- | | |
| 299 | − | | Required checks fail | Low | | |
| 300 | − | | It failed in the [merge queue](/guides/merge-queue/) | Low | | |
| 301 | − | | The reviewer agent asks for changes | Low | | |
| 302 | − | | A run was stopped at its cost or time cap | Low | | |
| 303 | − | | Sent back to revise | 1 point per revision, at most 3 | | |
| 304 | − | | Required checks have not finished, or have not run on its head | 1 point | | |
| 305 | − | | Checks passed only on a retry | 1 point | | |
| 306 | − | | The default branch has no required checks | 1 point | | |
| 307 | − | | No review yet, or the repository has no reviewer agent | 1 point | | |
| 308 | − | | The reviewer approved but left three or more comments on lines | 1 point | | |
| 309 | − | | Code changed and no test was added or changed | 1 point | | |
| 310 | − | | More than 400 lines changed; more than 1,000 | 1 point; 2 points | | |
| 311 | − | | More than 30 files changed | 1 point | | |
| 312 | − | | Files changed outside the area its [plan](/guides/outcomes/) expected; four or more | 1 point; 2 points | | |
| 313 | − | | Touches CI workflows, repository automation, secrets, infrastructure or `CODEOWNERS` | 2 points | | |
| 314 | − | | Its latest run used 80% or more of its cost or time cap | 1 point each | | |
| 315 | − | | Steps refused by [guardrails](/guides/guardrails/); three or more | 1 point; 2 points | | |
| 316 | − | | A question or handoff it sent another agent is unanswered | 2 points | | |
| 317 | − | | The agent said it was unsure about something | 1 point | | |
| 318 | − | ||
| 319 | − | At the end of every run that makes or revises a change, the agent is also | |
| 320 | − | asked how sure it is, and what it could not verify. g1t takes the lower of | |
| 321 | − | the two: what it observes can lower the agent's own word, never raise it. | |
| 322 | − | When the agent's word is lower, the reasons start with "agent says low", | |
| 323 | − | and the pull request lists what it was unsure about. | |
| 324 | − | ||
| 325 | − | For high confidence, the reasons say what it rests on: required checks pass, | |
| 326 | − | approved on the first review, tests added, a small change. | |
| 327 | − | ||
| 328 | − | The pull request's `confidence` in the | |
| 329 | − | [API](/reference/api/pull-requests/get-pull-request/) has the `level`, | |
| 330 | − | `reasons`, `self_reported`, `uncertain_about`, the `run_id` it was worked out | |
| 331 | − | after, and `assessed_at`. [Webhooks](/guides/webhooks/) for pull requests | |
| 332 | − | carry it too. | |
| 333 | − | ||
| 334 | − | ### Low-confidence changes wait for a person | |
| 335 | − | ||
| 336 | − | With **Ask a person before merging low-confidence changes** on, which it is | |
| 337 | − | unless someone turns it off, a g1t agent's change rated low is not merged | |
| 338 | − | by itself and does not join the merge queue, even with **Merge | |
| 339 | − | automatically when ready** on. Once everything else the repository asks | |
| 340 | − | for is met, it stops at **Needs you**, saying why, and Mission control | |
| 341 | − | lists it under **Needs you** with a **Low confidence** chip, the reasons in | |
| 342 | − | **What the agent already knows**, and the reasons again in **Why this | |
| 343 | − | needs you**. | |
| 344 | − | ||
| 345 | − | To let it land, approve it: a person's approval since the agent last | |
| 346 | − | revised lifts the hold, and it merges as the repository's rules say. To | |
| 347 | − | send it back, request changes. Merging it yourself works as usual. The | |
| 348 | − | setting is under **Settings → Branches and merging**, in **g1t agents**, | |
| 349 | − | and is `hold_low_confidence` in | |
| 350 | − | [`update_repo_settings`](/reference/api/repositories/update-repo-settings/). | |
| 351 | − | ||
| 352 | − | ## Choosing between pull requests | |
| 353 | − | ||
| 354 | − | Each pull request on the issue's page shows whether its checks passed. Open | |
| 355 | − | the ones that did, read their descriptions and changes, and merge the one | |
| 356 | − | you want. Merging lands it on `main` and closes the issue, which records | |
| 357 | − | that pull request as the one that resolved it. The other pull requests for | |
| 358 | − | the issue close as superseded. See | |
| 359 | − | [merging](/concepts/overview/#merging) for what happens when `main` has moved. | |
| 360 | − | ||
| 361 | − | ## Other things g1t agents do | |
| 362 | − | ||
| 363 | − | - **Review.** On a pull request that is ready, **Request review from g1t agent** | |
| 364 | − | has an agent read the change and post comments on lines, a summary and a | |
| 365 | − | verdict. | |
| 366 | − | - **Catch up.** When `main` has moved under a pull request, **Catch up with | |
| 367 | − | main** merges it in. When the two changed different files g1t does that | |
| 368 | − | itself in seconds, with no agent; otherwise an agent merges it in a | |
| 369 | − | sandbox and resolves any conflict | |
| 370 | − | ([catching up](/guides/pull-requests/#catching-up)). | |
| 371 | − | - **Finish a security update.** g1t raises a vulnerable dependency to its | |
| 372 | − | fixed version itself, with no agent. When raising the version is not | |
| 373 | − | enough (the bump fails, or the pull request's required checks fail | |
| 374 | − | because code must change), g1t opens an issue and puts g1t-agent on it. | |
| 375 | − | That session shows as **started by g1t** on the project's **Agents** page. | |
| 376 | − | See [security updates](/guides/security/#security-updates). | |
| 377 | − | ||
| 378 | − | Each runs in a sandbox of its own. | |
| 379 | − | ||
| 380 | − | ## Mentioning g1t-agent | |
| 381 | − | ||
| 382 | − | Write `@g1t-agent` in a comment on an issue or a pull request, with what | |
| 383 | − | you want, and it does it. The comment box offers to complete the name as | |
| 384 | − | you type `@`. | |
| 385 | − | ||
| 386 | − | | Where | You write | What happens | | |
| 387 | − | | --- | --- | --- | | |
| 388 | − | | An issue | A request: `@g1t-agent take this`, `@g1t-agent fix the empty case` | The issue is assigned to the agent, which opens a pull request, as if you had chosen **Assign**. | | |
| 389 | − | | An issue | A question: `@g1t-agent why does search time out?` | The agent reads the code on the default branch and answers in the thread. It changes nothing. | | |
| 390 | − | | A pull request g1t-agent made | A request: `@g1t-agent also handle the empty list` | The agent is sent back to make the change, with your comment as what to address, and the checks and review run again. If it is still working, it gets your comment as a message at its next step. | | |
| 391 | − | | Any pull request | `@g1t-agent review` | A review by a g1t agent, as with **Request review from g1t agent**. | | |
| 392 | − | | Any pull request | A question | The agent reads the change at its head and answers in the thread. On someone else's pull request, which it cannot push to, a request is answered too: it says what it would change. | | |
| 393 | − | ||
| 394 | − | A request is a comment whose words after the mention start with what to | |
| 395 | − | do (`take`, `fix`, `add`, `please rename`, `can you update`); a question | |
| 396 | − | starts with a question word or ends with a question mark. `review` near the | |
| 397 | − | start asks for a review. | |
| 398 | − | ||
| 399 | − | g1t-agent always replies in the thread, saying what it started or why it | |
| 400 | − | did not. Every run a mention starts shows on the project's **Agents** page | |
| 401 | − | as started by whoever mentioned it, and a mention that started nothing | |
| 402 | − | shows there as a failed run with the reason. | |
| 403 | − | ||
| 404 | − | **What does not count.** Mentions in code (`` `@g1t-agent` `` or a code | |
| 405 | − | block), in quoted lines (`> @g1t-agent …`), in email addresses | |
| 406 | − | (`ops@g1t-agent.dev`) and in longer names (`@g1t-agents`) are ignored. | |
| 407 | − | Matching ignores case. Agents mentioning `@g1t-agent` start nothing, so | |
| 408 | − | agents cannot set each other to work this way. | |
| 409 | − | ||
| 410 | − | **Who can.** People with the Write [role](/guides/access-and-roles/) or higher on the | |
| 411 | − | repository, members or not. Anyone else who mentions it gets a short reply | |
| 412 | − | saying that putting g1t-agent to work needs the Write role on the | |
| 413 | − | repository, and nothing starts. When | |
| 414 | − | the workspace's plan does not let the agent start (a free workspace with no | |
| 415 | − | trial left, a paused workspace, an issue at its spending cap), g1t-agent | |
| 416 | − | replies with why and where to fix it. When every agent slot is busy, it | |
| 417 | − | replies that the run is waiting for a free slot, and starts it when one | |
| 418 | − | finishes. | |
| 419 | − | ||
| 420 | − | Each comment starts one run at most; to ask again, write a new comment. | |
| 421 | − | ||
| 422 | − | ## The label rule | |
| 423 | − | ||
| 424 | − | Under a project's **Settings → Agents**, someone with the Maintain role or | |
| 425 | − | higher sets a label, such as `agent`. From then on, when someone with the | |
| 426 | − | Write role or higher gives an open issue that label, either | |
| 427 | − | when opening it or later, g1t-agent takes it: the issue is queued for an | |
| 428 | − | agent, the conversation says so, and the agent starts as soon as the | |
| 429 | − | project has room and nothing the issue depends on is still open, exactly as | |
| 430 | − | for a [plan's](/guides/outcomes/) issues. An issue that already had the | |
| 431 | − | label is not affected; removing and adding it again counts. **Turn off** | |
| 432 | − | removes the rule. | |
| 433 | − | ||
| 434 | − | ## Which model runs | |
| 435 | − | ||
| 436 | − | You do not pick one. You assign the work to `g1t-agent`, the way you would | |
| 437 | − | assign an issue to a colleague, and g1t routes it. The kind of work decides: | |
| 438 | − | ||
| 439 | − | | Work | Model today | | |
| 440 | − | | --- | --- | | |
| 441 | − | | Making a change for an issue, and revising it | Claude Sonnet 5.5 | | |
| 442 | − | | Reviewing a pull request | Claude Sonnet 5.5 | | |
| 443 | − | | Catching up with `main` and resolving conflicts | Claude Sonnet 5.5 | | |
| 444 | − | | Planning an outcome | Claude Sonnet 5.5 | | |
| 445 | − | ||
| 446 | − | Every session opens with a note naming the model that ran, and an agent's | |
| 447 | − | review says which model wrote it, so what you got is always on the record. | |
| 448 | − | When a better model for a kind of work appears, g1t changes the route and | |
| 449 | − | nothing you have set up needs to change. | |
| 450 | − | ||
| 451 | − | A pull request made by a g1t agent carries the label `g1t-agent`, and its | |
| 452 | − | commits are authored by `g1t agent`. | |
| 453 | − | ||
| 454 | − | ## How model traffic is routed | |
| 455 | − | ||
| 456 | − | g1t agents send model requests to g1t's model proxy at | |
| 457 | − | `https://models.g1t.sh`, with a token for their run in place of a key; see | |
| 458 | − | [your keys never reach a sandbox](/guides/models/#your-keys-never-reach-a-sandbox). | |
| 459 | − | Requests for g1t's hosted models go on through | |
| 460 | − | [Cloudflare AI Gateway](https://developers.cloudflare.com/ai-gateway/), | |
| 461 | − | which holds g1t's key. Each of those requests is tagged with the kind of | |
| 462 | − | work, the repository and the pull request, so spend can be read per pull | |
| 463 | − | request. Requests for a workspace's own provider go to that provider. | |
| 464 | − | ||
| 465 | − | If you run your own copy of g1t, these settings control it: | |
| 466 | − | ||
| 467 | − | | Setting | Where | What it does | | |
| 468 | − | | --- | --- | --- | | |
| 469 | − | | `AGENT_ROUTES` | Runner | The model for each kind of work: `implement`, `review`, `update` and `plan`. | | |
| 470 | − | | `MODELS_URL` | Runner | Where sandboxes send model requests: the model proxy. | | |
| 471 | − | | `AI_GATEWAY_ID` | Model proxy | The gateway hosted requests go through. Empty sends them to the provider directly. | | |
| 472 | − | | `AI_GATEWAY_TOKEN` | Model proxy | Secret. Authenticates to the gateway. | | |
| 473 | − | | `ANTHROPIC_API_KEY` | Model proxy | Secret. The provider's key, if the gateway does not hold it. | | |
| 474 | − | ||
| 475 | − | ## What it costs | |
| 476 | − | ||
| 477 | − | A workspace pays for the g1t agents that work on its repositories, after | |
| 478 | − | they run: each run is charged its sandbox by the second, at cost plus 20%, | |
| 479 | − | and, on g1t's hosted models, what AI Gateway priced its model requests at, | |
| 480 | − | plus 20%. A workspace's [own provider](/guides/models/) bills it for the | |
| 481 | − | model directly. See | |
| 482 | − | [Usage and billing](/guides/usage-and-billing/) for how prices are set and | |
| 483 | − | the limits on usage not yet paid for. | |
| 484 | − | The workspace's **Usage** page shows what its agents have cost, by day, | |
| 485 | − | kind of work, repository, model and pull request. See | |
| 486 | − | [usage and billing](/guides/usage-and-billing/). | |
| 487 | − | ||
| 488 | − | ## What a sandbox has | |
| 489 | − | ||
| 490 | − | Git, common shell tools, and toolchains for Node.js, Python, Go and Rust, so | |
| 491 | − | an agent can build and test most projects. If your project needs something | |
| 492 | − | else, the agent will say in its summary what it could not run. | |
| 493 | − | ||
| 494 | − | A workspace (or a project) can send its agents' work to | |
| 495 | − | [its own runners](/guides/self-hosted-runners/#agents-on-your-runners) | |
| 496 | − | instead, so agents build and test with what those machines have. The agent | |
| 497 | − | works the same way there, with the same short-lived credentials, and its | |
| 498 | − | model calls still go through g1t; the machine time is free. g1t's network | |
| 499 | − | guardrails cannot be enforced on your machines, and the run says so. | |
| 500 | − | ||
| 501 | − | ## Who can run agents | |
| 502 | − | ||
| 503 | − | Putting an agent to work (assigning it, mentioning it, asking it for a | |
| 504 | − | review, planning, sending it back to revise) needs the Write | |
| 505 | − | [role](/guides/access-and-roles/) or higher on the repository. That | |
| 506 | − | includes an [outside collaborator](/guides/access-and-roles/#outside-collaborators) | |
| 507 | − | with Write: their runs are charged to the repository's workspace, as a | |
| 508 | − | member's are, and count against its plan, caps and agent slots. They see | |
| 509 | − | what their agents do, but not which model ran or what a run cost; those | |
| 510 | − | are for members of the workspace. A run for an outside collaborator is | |
| 511 | − | told the project's memory, never the workspace's. | |
| 512 | − | ||
| 513 | − | Agents cost g1t real money, so they run for paid workspaces. A free | |
| 514 | − | workspace has the whole forge, and two ways to try agents: | |
| 515 | − | ||
| 516 | − | - **The trial.** $5 of usage, once per workspace, after a card check. | |
| 517 | − | - **The open-source pool.** Workflows and the merge queue on public | |
| 518 | − | repositories, after the same card check. It does not pay for agents. | |
| 519 | − | ||
| 520 | − | This holds whether the agent uses g1t's hosted models or | |
| 521 | − | [the workspace's own model provider](/guides/models/): the sandbox an agent | |
| 522 | − | works in is g1t's either way. Before anyone assigns, asks for a review or | |
| 523 | − | plans, a free workspace's pages say "Agents need a paid workspace or the | |
| 524 | − | free trial", with a link to its **Billing** page. | |
| 525 | − | ||
| 526 | − | Every agent run, of every kind (making a change, revising, reviewing, | |
| 527 | − | catching up, planning, answering a mention), asks billing before it | |
| 528 | − | starts. Billing reserves what the run is expected to cost: its model's | |
| 529 | − | recent average (about $0.10 to make a change or plan, $0.07 to review) | |
| 530 | − | plus its sandbox for its whole time cap. When the run ends, what it really | |
| 531 | − | cost is settled against that. If billing refuses, nothing starts, and you | |
| 532 | − | see why where you started it: | |
| 533 | − | ||
| 534 | − | | Where you started it | Where the refusal shows | | |
| 535 | − | | --- | --- | | |
| 536 | − | | **Assign to g1t agent**, **Request review**, **Plan it**, catching up | Under the button | | |
| 537 | − | | A mention or the label rule | A comment from g1t-agent on the issue or pull request | | |
| 538 | − | | A step g1t takes by itself (a review, a revision, a catch-up) | The pull request's status, which then waits for you | | |
| 539 | − | ||
| 540 | − | Each refusal says what to do: start the plan or the trial, raise the spend | |
| 541 | − | limit, or wait for next month's open-source pool, with the page to do it | |
| 542 | − | on. | |
| 543 | − | ||
| 544 | − | ### Caps on a plan | |
| 545 | − | ||
| 546 | − | A workspace's plan sets caps on its agents. A new paid workspace in its | |
| 547 | − | first month, and a workspace on the trial, has tighter ones. The amounts | |
| 548 | − | are on [usage and billing](/guides/usage-and-billing/#caps). | |
| 549 | − | ||
| 550 | − | | Cap | What happens at it | | |
| 551 | − | | --- | --- | | |
| 552 | − | | Agents at once | A run over the cap waits for a free slot instead of being refused. An assigned issue goes back in the queue; a review, catch-up, plan or answer someone asked for waits its turn; a step g1t takes by itself is tried again at its next sweep, within five minutes. Each says "Waiting for a free slot". | | |
| 553 | − | | Time per run | The lower of the project's [guardrails](/guides/guardrails/) time cap for that kind of run and the plan's. | | |
| 554 | − | | Cost per run | The lower of the guardrails' cost cap and the plan's. The agent is stopped when it reaches it, as with any cost cap. | | |
| 555 | − | | Cost per issue | What every agent run on an issue and its pull requests has cost in all. Past it, g1t-agent does not start on that issue again and says so on it; an owner can raise the cap on the **Billing** page. | | |
| 556 | − | ||
| 557 | − | When the workspace's compute is paused (a spend spike waiting for an owner, | |
| 558 | − | or a hold by g1t), nothing new starts, and the refusal gives the reason. | |
| 559 | − | ||
| 560 | − | ### When billing cannot be reached | |
| 561 | − | ||
| 562 | − | g1t's own billing service could be briefly unreachable. Then: | |
| 563 | − | ||
| 564 | − | - A paid workspace's runs go ahead, and the miss is logged. A billing blip | |
| 565 | − | never stops a paying customer's work. | |
| 566 | − | - A free workspace's runs do not start: they would be paid for by nobody. | |
| 567 | − | Try again in a minute. | |
| 568 | − | ||
| 569 | − | g1t decides which a workspace is from its plan, or from the last plan it | |
| 570 | − | saw for it in the past day. | |
| 571 | − | ||
| 572 | − | ### Other limits | |
| 573 | − | ||
| 574 | − | - A run has two hours. After that its credentials expire and it can no | |
| 575 | − | longer push or report. | |
| 576 | − | - Agents do not run on an [archived](/guides/managing-repositories/#archive-a-repository) | |
| 577 | − | or deleted repository: nothing new starts, whether from an assignment, | |
| 578 | − | a mention or the label rule, and a run under way cannot push or merge. | |
| 579 | − | What was refused does not start by itself when the repository is | |
| 580 | − | unarchived or restored; assign the work again. | |
| 581 | − | ||
| 582 | − | ## Credentials | |
| 583 | − | ||
| 584 | − | Every sandbox run gets credentials of its own, made when it starts and | |
| 585 | − | revoked the moment it stops. They are not your access tokens, and they are | |
| 586 | − | not listed with them. | |
| 587 | − | ||
| 588 | − | Each credential carries a composite identity: the agent, acting on behalf | |
| 589 | − | of the person who started the work. A run you started by assigning an issue | |
| 590 | − | is `g1t-agent on behalf of you`, and that is how it appears in the | |
| 591 | − | [audit log](/guides/audit-log/), on the run's page and in the pull | |
| 592 | − | request's **Agent** panel. | |
| 593 | − | ||
| 594 | − | What it may do is the intersection of two things: | |
| 595 | − | ||
| 596 | − | - **The run's scope.** The credential is bound to the run, its repository, | |
| 597 | − | and what that kind of run needs. It expires no later than the run's | |
| 598 | − | timeout. | |
| 599 | − | - **What you may do now.** It works only in the repository's workspace, | |
| 600 | − | with your [role](/guides/access-and-roles/) on the repository as it is now, and never | |
| 601 | − | more than Write: an owner's agent has Write, not Admin. If you leave the | |
| 602 | − | workspace or lose your role, every agent working on your behalf there | |
| 603 | − | loses it too. It can never change who has access. | |
| 604 | − | ||
| 605 | − | A sandbox holds two credentials. One is for g1t's runner, which clones, | |
| 606 | − | pushes the result and records the session; downstream it acts as you, so | |
| 607 | − | what it pushes is yours, within the run's scope. The other is for the | |
| 608 | − | agent's own tools over MCP, and acts as the agent; it cannot be used with | |
| 609 | − | git at all. | |
| 610 | − | ||
| 611 | − | | Kind of run | Git | API and MCP tools | | |
| 612 | − | | --- | --- | --- | | |
| 613 | − | | Implement | Reads the repository; pushes to its pull request's fork only | Records the session and marks its own pull request ready; tools to read issues, pull requests, the merge queue, workflow runs and memory, to [search all of g1t](/guides/search/) and the workspace's context hub, open issues, comment, remember, and message other agents | | |
| 614 | − | | Revise, answer | Reads the repository; pushes to the pull request's fork, or to its branch only when the change is a branch of the repository | Records the session of its own pull request; the same tools as implement | | |
| 615 | − | | Catch up | Reads the repository; pushes to the pull request's fork or branch only | Records the session of its own pull request | | |
| 616 | − | | Review | Reads the change and the repository; pushes nothing | Reports its review through its own run | | |
| 617 | − | | Plan | Reads the repository; pushes nothing | Reports its plan through its own run, for a person to apply; it can create issues in its repository only | | |
| 618 | − | | Merge check | Reads the change; pushes nothing | None | | |
| 619 | − | | Merge queue | Reads each queued change; pushes the queue's own branch only | None | | |
| 620 | − | | Deploy | Reads the commit it builds; pushes nothing | None | | |
| 621 | − | ||
| 622 | − | Nothing an agent's credential holds can reach another repository, or a | |
| 623 | − | workspace's settings, members, access tokens, billing, integrations, | |
| 624 | − | webhooks, secrets and variables, or workflows' controls. It cannot merge a | |
| 625 | − | pull request or put more agents to work. It cannot change its repository's | |
| 626 | − | details or default branch, rename it or its branches, make it public or | |
| 627 | − | private, archive, transfer, delete, restore or purge it; see | |
| 628 | − | [managing a repository](/guides/managing-repositories/). A call that would is refused, and | |
| 629 | − | the refusal is recorded with the rule that refused it: | |
| 630 | − | ||
| 631 | − | | Rule | Refused because | | |
| 632 | − | | --- | --- | | |
| 633 | − | | `never` | No agent's credential may ever do this. | | |
| 634 | − | | `scope:operation` | The run's kind does not include this operation. | | |
| 635 | − | | `scope:repository` | It names a repository other than the run's. | | |
| 636 | − | | `scope:pull` | The runner tried to change a pull request other than its own. | | |
| 637 | − | | `on-behalf-of:membership` | The person the agent works for is no longer a member of the workspace, and has no role on its repositories. | | |
| 638 | − | | `git:read`, `git:push`, `git:ref` | The run has no grant to clone that repository, push to it, or move that branch or tag. | | |
| 639 | − | | `git:not-a-run` | An agent's tools credential was used with git. | | |
| 640 | − | ||
| 641 | − | Personal and workspace access tokens are unchanged by any of this. | |
| 642 | − | ||
| 643 | − | ## What a sandbox can reach | |
| 644 | − | ||
| 645 | − | A sandbox holds one fork and its run's credentials, which expire when the | |
| 646 | − | run's time is up and are revoked as soon as it stops. Those credentials, | |
| 647 | − | and the model key the agent runs on, are removed from anything recorded in | |
| 648 | − | the session. | |
| 649 | − | ||
| 650 | − | ## Guardrails | |
| 651 | − | ||
| 652 | − | A workspace decides what its agents may do in their sandboxes, and each | |
| 653 | − | project can override it: which hosts a sandbox can reach (g1t, the package | |
| 654 | − | registries the project needs, and domains you list; enforced outside the | |
| 655 | − | sandbox), which commands the harness refuses (force-pushing, rewriting the | |
| 656 | − | default branch, reading outside the project, printing the environment, | |
| 657 | − | sudo, and your own patterns), and how much one run may cost and how long it | |
| 658 | − | may take. A run that is refused something shows it as a step; one that | |
| 659 | − | reaches a cap is stopped and its pull request waits for you. A run on a | |
| 660 | − | fork's head loads none of the fork's `CLAUDE.md`, `.claude` settings, | |
| 661 | − | hooks, MCP servers or commands. See [guardrails](/guides/guardrails/) for | |
| 662 | − | every rule and exactly how each is enforced. |
| 116 | 116 | https://g1t.sh/pulls/<pull request id>.git | |
| 117 | 117 | ``` | |
| 118 | 118 | ||
| 119 | − | Only whoever opened the pull request can push to it. Pushes to a fork | |
| 119 | + | Only whoever opened the pull request can push to it, or, for one g1t | |
| 120 | + | made, whoever asked for it. Pushes to a fork | |
| 120 | 121 | update the pull request's head commit on its page. | |
| 121 | 122 | ||
| 122 | 123 | ## Limits |
| 196 | 196 | - The sandbox's credentials. The agent itself holds none of g1t's: the | |
| 197 | 197 | runner clones and pushes with credentials passed per command, and pushes | |
| 198 | 198 | only to the run's own fork or branch, never with force. See | |
| 199 | − | [credentials](/guides/g1t-agents/#credentials). | |
| 199 | + | [credentials](/guides/working-with-g1t/#credentials). | |
| 200 | 200 | - Branch protection on the repository, which g1t enforces when a push | |
| 201 | 201 | arrives, whatever the sandbox did. | |
| 202 | 202 | ||
| 203 | 203 | ### Changes from forks | |
| 204 | 204 | ||
| 205 | 205 | When a run checks out a fork's head (revising, reviewing or answering on | |
| 206 | − | any pull request from a fork, including g1t-agent's own, and replying to a | |
| 206 | + | any pull request from a fork, including g1t's own, and replying to a | |
| 207 | 207 | mention on one), the harness loads nothing from that checkout: no | |
| 208 | 208 | `CLAUDE.md`, no `.claude/settings.json` or `settings.local.json` (so none | |
| 209 | 209 | of their hooks or permissions), no `.mcp.json` servers, and no commands or | |
| 210 | 210 | skills. g1t's guardrails, its tools and the repository's instructions from | |
| 211 | 211 | its own branches still apply. The run's session says so at the start. This | |
| 212 | 212 | follows the rule g1t uses for | |
| 213 | − | [repository instructions](/guides/g1t-agents/#repository-instructions): | |
| 213 | + | [repository instructions](/guides/working-with-g1t/#repository-instructions): | |
| 214 | 214 | they are read from the repository's own branches, never from a fork. | |
| 215 | 215 | ||
| 216 | 216 | ## Caps | |
| 244 | 244 | stops at $2 unless an owner raises it (see | |
| 245 | 245 | [caps](/guides/usage-and-billing/#caps)). A new paid workspace's first | |
| 246 | 246 | month, and the trial, also cap every run's time at 60 minutes (see | |
| 247 | − | [who can run agents](/guides/g1t-agents/#who-can-run-agents)). A run gets | |
| 247 | + | [who can run agents](/guides/working-with-g1t/#who-can-run-agents)). A run gets | |
| 248 | 248 | the lower of its guardrails' cap and its plan's, for time and for cost, | |
| 249 | 249 | and its page shows the cap it got. | |
| 250 | 250 |
| 32 | 32 | linking the new occurrence. | |
| 33 | 33 | - **A recovery** (Datadog) is noted on the issue. | |
| 34 | 34 | ||
| 35 | − | Turn on **Put a g1t agent on each new issue** and an agent starts on the | |
| 35 | + | Turn on **Assign each new issue to g1t** and g1t starts on the | |
| 36 | 36 | issue as soon as it opens: it makes the change, is reviewed, revises, and | |
| 37 | 37 | lands through your repository's rules, often before anyone has looked. A | |
| 38 | 38 | reopened issue gets an agent again. Agents need a way to reach a model: | |
| 138 | 138 | ||
| 139 | 139 | ### Agents read tickets the work mentions | |
| 140 | 140 | ||
| 141 | − | When a g1t agent starts on an issue, or plans an outcome, g1t looks for | |
| 141 | + | When g1t starts on an issue, or plans an outcome, g1t looks for | |
| 142 | 142 | ticket keys and addresses in the text (`TECH-1234`, | |
| 143 | 143 | `https://acme.atlassian.net/browse/TECH-1234`, | |
| 144 | 144 | `https://linear.app/acme/issue/ENG-42/…`, a Sentry issue's address) and |
| 40 | 40 | ||
| 41 | 41 | Renaming the repository or its default branch, changing its visibility | |
| 42 | 42 | and archiving it are done by a person, not a workspace token. | |
| 43 | − | A g1t agent's token can never use the tools that make these changes, | |
| 44 | − | whatever its run. See [credentials](/guides/g1t-agents/#credentials). | |
| 43 | + | g1t's token can never use the tools that make these changes, | |
| 44 | + | whatever its run. See [credentials](/guides/working-with-g1t/#credentials). | |
| 45 | 45 | ||
| 46 | 46 | ## Edit the details | |
| 47 | 47 | ||
| 374 | 374 | | Response | Why | What to do | | |
| 375 | 375 | | --- | --- | --- | | |
| 376 | 376 | | `401 unauthenticated` | No token, or one that is not valid. | Send a personal access token. | | |
| 377 | − | | `403 forbidden` | You are not an owner, for a change that needs one; or a g1t agent's token was used. | Ask an owner of the workspace. | | |
| 377 | + | | `403 forbidden` | You are not an owner, for a change that needs one; or g1t's token was used. | Ask an owner of the workspace. | | |
| 378 | 378 | | `404 not_found` | No such repository or branch, or you cannot see it. For restore and purge: no deleted repository had that path, or it was purged. | Check the path. A deleted repository is named by the path it had. | | |
| 379 | 379 | | `409 conflict` | The new name is taken in the workspace, by a repository or a recently deleted one; or a branch with the new name exists. | Pick another name, or purge the deleted repository first. | | |
| 380 | 380 | | `422 invalid` | `confirm` is not the repository's full name; the name, branch name or website is not valid; or the new default branch does not exist. | Type `<workspace>/<repo>` exactly; push the branch first. | |
| 139 | 139 | file, which opens in the pull request's changes. | |
| 140 | 140 | 3. The entries that were tested on top of it are tested again without it. | |
| 141 | 141 | ||
| 142 | − | A g1t agent's pull request is then sent back to revise, as for any failed | |
| 142 | + | A pull request g1t opened is then sent back to revise, as for any failed | |
| 143 | 143 | check, starting from `main` as it is now. The revision counts towards | |
| 144 | 144 | **Revisions before asking you**. Once it is ready again, a repository with | |
| 145 | 145 | **Merge automatically when ready** on adds it to the queue again by itself; | |
| 183 | 183 | { | |
| 184 | 184 | "number": 44, | |
| 185 | 185 | "title": "Add a --shout flag", | |
| 186 | − | "agent": "g1t-agent", | |
| 186 | + | "agent": "g1t", | |
| 187 | 187 | "state": "testing", | |
| 188 | 188 | "ahead": [41], | |
| 189 | 189 | "base_commit": "8f3c2e1…", |
| 76 | 76 | ||
| 77 | 77 | Each can go to g1t's models, or to any of your providers on any of its | |
| 78 | 78 | models. An Anthropic provider also offers **g1t's choice of Claude**, | |
| 79 | − | which runs g1t's pick for that kind of work on your key. For example: make | |
| 79 | + | which runs g1t's large-tier model, Claude Sonnet 5.5 today, on your key. | |
| 80 | + | Routing between tiers by the size of the work is only for g1t's hosted | |
| 81 | + | models; see [which model runs](/guides/working-with-g1t/#which-model-runs). For example: make | |
| 80 | 82 | changes on Claude through your Anthropic key, review on GPT through your | |
| 81 | 83 | OpenAI key, and catch up on a small model through OpenRouter. | |
| 82 | 84 |
| 1 | 1 | --- | |
| 2 | 2 | title: Hand off an outcome | |
| 3 | − | description: Write what should be true, let an agent plan the issues, and follow g1t agents as they land them. | |
| 3 | + | description: Write what should be true, let an agent plan the issues, and follow g1t as it lands them. | |
| 4 | 4 | --- | |
| 5 | 5 | ||
| 6 | 6 | You do not have to split work into issues yourself. Write the outcome you | |
| 7 | 7 | want on a repository's **Plan** page. An agent reads the repository and | |
| 8 | 8 | proposes the issues that would get there, with what done means for each and | |
| 9 | 9 | the order they have to land in. You read the plan, keep what you want, and open | |
| 10 | − | it. g1t agents then work on the issues, as many at once as the dependencies | |
| 10 | + | it. g1t then works on the issues, as many at once as the dependencies | |
| 11 | 11 | allow, and the outcome page shows each one until it lands. | |
| 12 | 12 | ||
| 13 | − | Planning and g1t agents work in any workspace that | |
| 14 | − | [can run agents](/guides/g1t-agents/#who-can-run-agents) (the plan or the | |
| 13 | + | Planning and g1t's agent work in any workspace that | |
| 14 | + | [can run agents](/guides/working-with-g1t/#who-can-run-agents) (the plan or the | |
| 15 | 15 | trial) and has a model: [its own model provider](/guides/models/), or g1t's | |
| 16 | 16 | hosted models where they are open to it. The agents' runs are charged to | |
| 17 | 17 | the workspace; see | |
| 59 | 59 | ||
| 60 | 60 | | Choice | What happens | | |
| 61 | 61 | | --- | --- | | |
| 62 | − | | **Open these and assign g1t agents** | The issues are opened and queued for g1t agents. Agents start at once on every issue that depends on nothing, working in parallel, and on the others as what they depend on merges. | | |
| 62 | + | | **Open these and assign g1t** | The issues are opened and queued for g1t. It starts at once on every issue that depends on nothing, working in parallel, and on the others as what they depend on merges. | | |
| 63 | 63 | | **Only open the issues** | The issues are opened, each blocked by the ones it depends on. Nobody is put to work on them. | | |
| 64 | 64 | ||
| 65 | 65 | A dependency on an issue you unticked is dropped with it. A plan is applied | |
| 67 | 67 | ||
| 68 | 68 | ### How queued issues start | |
| 69 | 69 | ||
| 70 | − | An issue queued for a g1t agent starts when: | |
| 70 | + | An issue queued for g1t starts when: | |
| 71 | 71 | ||
| 72 | 72 | - every issue it depends on has closed, normally because a pull request for | |
| 73 | 73 | it merged; and | |
| 74 | − | - the repository has room. At most six g1t agents make changes in one | |
| 74 | + | - the repository has room. g1t makes at most six changes in one | |
| 75 | 75 | repository at once. The rest wait their turn, which also leaves sandboxes | |
| 76 | 76 | free for reviews. | |
| 77 | 77 | ||
| 78 | 78 | Each queued issue says so in its conversation, for example "queued this for | |
| 79 | − | g1t-agent, to start once #41 has merged". From there each issue is | |
| 80 | − | [seen through](/guides/g1t-agents/#seeing-it-through) like any other a g1t | |
| 81 | − | agent works on: checks, review, revision, and merging under the | |
| 79 | + | g1t, to start once #41 has merged". From there each issue is | |
| 80 | + | [seen through](/guides/working-with-g1t/#seeing-it-through) like any other g1t | |
| 81 | + | works on: checks, review, revision, and merging under the | |
| 82 | 82 | repository's rules. | |
| 83 | 83 | ||
| 84 | 84 | ## Follow the outcome | |
| 105 | 105 | | Blocked | Waiting for the issues it depends on to land. | | |
| 106 | 106 | | Waiting for an agent | Queued, and waiting for an agent to be free. | | |
| 107 | 107 | | Open | Nobody is working on it. | | |
| 108 | − | | Agent working | A g1t agent is making the change. | | |
| 108 | + | | Agent working | g1t is making the change. | | |
| 109 | 109 | | Checking | Its workflows are running, or a required check has not reported yet. | | |
| 110 | − | | In review | A g1t agent is reviewing the change. | | |
| 110 | + | | In review | g1t is reviewing the change. | | |
| 111 | 111 | | Revising | The agent was sent back by the checks, a review or a person. | | |
| 112 | 112 | | Catching up | The agent is merging in the branch it will land on, which has moved. | | |
| 113 | 113 | | In the merge queue | It is being tested with the changes ahead of it. See [the merge queue](/guides/merge-queue/). | | |
| 151 | 151 | curl https://api.g1t.sh/repos/acme/greeter/plans/pln_01… \ | |
| 152 | 152 | -H "Authorization: Bearer $G1T_TOKEN" | |
| 153 | 153 | ||
| 154 | − | # 3. Open issues 1 and 3 and put g1t agents on them. | |
| 154 | + | # 3. Open issues 1 and 3 and assign them to g1t. | |
| 155 | 155 | curl -X POST https://api.g1t.sh/repos/acme/greeter/plans/pln_01…/apply \ | |
| 156 | 156 | -H "Authorization: Bearer $G1T_TOKEN" \ | |
| 157 | 157 | -H "Content-Type: application/json" \ |
| 70 | 70 | | Deploy to production | A production build has gone live. | Deployments settings, or **Deployments** once they are on. | | |
| 71 | 71 | | Add a custom domain | The project has a [custom domain](/guides/deployments/#custom-domains). | Domain settings. | | |
| 72 | 72 | | Open a preview | A branch or pull request has had a [preview](/guides/deployments/#previews-of-branches). | A new pull request. | | |
| 73 | − | | Set up repository instructions | `AGENTS.md` or `CLAUDE.md` is at the root of the default branch. | The instructions on the **Agents** page. See [repository instructions](/guides/g1t-agents/#repository-instructions). | | |
| 74 | − | | Assign a first issue to g1t-agent | g1t-agent has had a run, a pull request or an issue here. | A new issue. | | |
| 73 | + | | Set up repository instructions | `AGENTS.md` or `CLAUDE.md` is at the root of the default branch. | The instructions on the **Agents** page. See [repository instructions](/guides/working-with-g1t/#repository-instructions). | | |
| 74 | + | | Assign a first issue to g1t | g1t has had a run, a pull request or an issue here. | A new issue. | | |
| 75 | 75 | ||
| 76 | 76 | The card goes away when every step is done. To hide it sooner, choose | |
| 77 | 77 | **×** on it. That hides it for this project in this browser only. |
| 8 | 8 | whether it merges cleanly, and the button that merges it. Everything in it | |
| 9 | 9 | updates by itself while something is still running. | |
| 10 | 10 | ||
| 11 | + | A pull request g1t made shows **g1t** as its author and **requested by** | |
| 12 | + | the person who asked for it. That person can manage it as its author could, | |
| 13 | + | and cannot approve it; see | |
| 14 | + | [who a pull request is for](/guides/working-with-g1t/#who-a-pull-request-is-for). | |
| 15 | + | ||
| 11 | 16 | ## Checks | |
| 12 | 17 | ||
| 13 | 18 | A pull request's checks are the statuses reported on its head commit. Most | |
| 66 | 71 | | Require a pull request to change the default branch | Off | Refuses pushes to the default branch; changes reach it only by merging. See [protected branches](/guides/git/#protected-branches). | | |
| 67 | 72 | | Required status checks | None | The checks that must pass on a pull request's head before it merges. | | |
| 68 | 73 | | Required approvals | None | How many reviewers must approve before a merge, 0 to 3 on the page (up to 6 from the API). A reviewer who asked for changes blocks it. | | |
| 69 | − | | A g1t agent's approval counts | On | Off means approvals have to come from people. | | |
| 74 | + | | g1t's approval counts | On | Off means approvals have to come from people. | | |
| 70 | 75 | | Require branches to be up to date before merging | Off | On means a pull request behind the default branch has to catch up, and its checks run again, before it merges. | | |
| 71 | 76 | | Merge through a queue | Off | See [merge queue](/guides/merge-queue/). | | |
| 72 | 77 | | Allow bypassing required checks | On | Lets someone who may merge tick **bypass** when merging, to merge without the required checks passing. Off means nobody can. | | |
| 90 | 95 | ||
| 91 | 96 | The same rule holds wherever a pull request merges: the merge button, | |
| 92 | 97 | [`merge_pull_request`](/reference/api/pull-requests/merge-pull-request/), | |
| 93 | − | a g1t agent's [automatic merge](/guides/g1t-agents/#merging-automatically) | |
| 98 | + | a g1t agent's [automatic merge](/guides/working-with-g1t/#merging-automatically) | |
| 94 | 99 | and the [merge queue](/guides/merge-queue/). With **Allow bypassing required | |
| 95 | 100 | checks** on, the merge button has a **bypass** box, and the API takes | |
| 96 | 101 | `ignore_checks: true`. | |
| 146 | 151 | be resolved**, the conflicting files, each linked to its diff, and three | |
| 147 | 152 | ways to resolve them: | |
| 148 | 153 | ||
| 149 | − | - **Resolve with g1t agent.** An agent merges the target branch in, | |
| 154 | + | - **Resolve with g1t.** g1t merges the target branch in, | |
| 150 | 155 | resolves the conflicts keeping what both sides meant, and pushes the | |
| 151 | 156 | result. It is told which files conflict. Available to whoever can push to | |
| 152 | − | the pull request: for a pull request's fork, whoever opened it; for a | |
| 153 | − | branch, anyone with the Write role or higher. | |
| 157 | + | the pull request: for a pull request's fork, whoever opened it (whoever | |
| 158 | + | asked g1t for one it made); for a branch, anyone with the Write role or | |
| 159 | + | higher. | |
| 154 | 160 | - **Resolve in the browser.** Coming soon. | |
| 155 | 161 | - **On the command line.** The box lists the commands, each with a copy | |
| 156 | 162 | button. For a pull request from a branch: | |
| 184 | 190 | ||
| 185 | 191 | When the target branch has moved, the merge box says **main has moved since | |
| 186 | 192 | this was made**. Whoever can push to the pull request (whoever opened it, | |
| 187 | − | for one in its own fork; anyone with the Write [role](/guides/access-and-roles/) or higher, | |
| 188 | − | for a branch) can | |
| 193 | + | or asked g1t for it, for one in its own fork; anyone with the Write | |
| 194 | + | [role](/guides/access-and-roles/) or higher, for a branch) can | |
| 189 | 195 | press **Catch up with main now**: | |
| 190 | 196 | ||
| 191 | 197 | 1. **When the two changed different files**, g1t merges `main` in itself, | |
| 195 | 201 | up to date with main**, and the workflows run again on the | |
| 196 | 202 | new commit, as after any push. | |
| 197 | 203 | 2. **When both changed some of the same files**, a sandbox merges `main` in | |
| 198 | − | with git, and a [g1t agent](/guides/g1t-agents/) resolves any conflict. | |
| 199 | − | The box says what is happening (**g1t-agent is resolving conflicts with | |
| 204 | + | with git, and [g1t](/guides/working-with-g1t/) resolves any conflict. | |
| 205 | + | The box says what is happening (**g1t is resolving conflicts with | |
| 200 | 206 | main** when the merge is known to conflict) with the run's live step and | |
| 201 | 207 | how long it has taken. It usually takes about a minute. When the result | |
| 202 | 208 | is pushed, the box shows the pull request up to date; if the run fails, | |
| 207 | 213 | where it was when the catch-up started. If someone pushed to it meanwhile, | |
| 208 | 214 | the catch-up stops with nothing lost, and you can press it again. | |
| 209 | 215 | ||
| 210 | − | The second case needs g1t agents enabled for the workspace and is | |
| 216 | + | The second case needs g1t's agent enabled for the workspace and is | |
| 211 | 217 | [charged](/guides/usage-and-billing/#what-is-charged) as agent work; the | |
| 212 | 218 | first is not. | |
| 213 | 219 | ||
| 214 | − | A [g1t agent's](/guides/g1t-agents/) pull request that is found to conflict | |
| 220 | + | A pull request [g1t](/guides/working-with-g1t/) opened that is found to conflict | |
| 215 | 221 | is sent back to resolve it by itself, before it is ready. | |
| 216 | 222 | ||
| 217 | 223 | ## From the API |
| 70 | 70 | | `is:open`, `is:closed` | Issues and pull requests by state | Issues, pull requests | | |
| 71 | 71 | | `is:merged`, `is:draft` | Pull requests that were merged, or are still drafts | Pull requests | | |
| 72 | 72 | | `is:public`, `is:private` | Results from public or private repositories | Repositories, code, issues, pull requests | | |
| 73 | − | | `author:ana` | Opened by someone | Issues, pull requests | | |
| 73 | + | | `author:ana` | Opened by someone. What g1t opened is `author:g1t`, and shows who it was for | Issues, pull requests | | |
| 74 | 74 | | `label:bug` | With a label. Quote a label with spaces: `label:"good first issue"` | Issues, pull requests | | |
| 75 | 75 | | `type:code` | Which tab to open: `repositories`, `code`, `issues`, `pulls` or `people` | All | | |
| 76 | 76 |
| 78 | 78 | - pushes, schedules, manual runs and the merge queue; | |
| 79 | 79 | - pull requests whose author has the Write [role](/guides/access-and-roles/) | |
| 80 | 80 | or higher on the repository, a member or an | |
| 81 | − | [outside collaborator](/guides/access-and-roles/#outside-collaborators); | |
| 82 | − | - pull requests from g1t's agents. | |
| 81 | + | [outside collaborator](/guides/access-and-roles/#outside-collaborators). | |
| 82 | + | For one g1t made, that is whoever asked for it; | |
| 83 | + | - pull requests g1t opened on its own, such as a security update. | |
| 83 | 84 | ||
| 84 | 85 | Anyone else's pull request, such as one from a fork or by someone with | |
| 85 | 86 | Read or Triage (who may open one on a private repository too), runs its |
| 292 | 292 | Install scripts never run. A lockfile the sandbox cannot change this way, | |
| 293 | 293 | such as a `requirements.txt` that pins with `--hash`, is handled as | |
| 294 | 294 | [when code has to change](#when-code-has-to-change): if no branch has been | |
| 295 | − | pushed 45 minutes after the update started, g1t-agent gets an issue for it. | |
| 295 | + | pushed 45 minutes after the update started, g1t gets an issue for it. | |
| 296 | 296 | ||
| 297 | 297 | ### When code has to change | |
| 298 | 298 | ||
| 299 | 299 | Sometimes raising the version is not enough: the bump fails, or the pull | |
| 300 | 300 | request's required checks fail because code must change. Then g1t opens an | |
| 301 | − | issue and puts [g1t-agent](/guides/g1t-agents/) on it. That agent's session | |
| 301 | + | issue and assigns it to [g1t](/guides/working-with-g1t/). That session | |
| 302 | 302 | shows as **started by g1t**. This is the only time an agent is involved in | |
| 303 | 303 | a security update. | |
| 304 | 304 |
| 170 | 170 | ||
| 171 | 171 | **Settings → Runners → Where work runs** can send g1t's own work to your | |
| 172 | 172 | runners too: agent runs, checks, reviews, merge checks and the merge queue. | |
| 173 | − | Switch on **Run g1t-agent work on self-hosted runners** and give the labels | |
| 173 | + | Switch on **Run g1t's work on self-hosted runners** and give the labels | |
| 174 | 174 | a runner needs to take it (`self-hosted` is always one). | |
| 175 | 175 | ||
| 176 | 176 | - The agent works exactly as in g1t's sandbox: the same harness, with a |
| 23 | 23 | | A status page of your own | Works, at `http://localhost:8788` ([below](#the-status-page)) | | |
| 24 | 24 | | Webhooks, integrations | Run, but scheduled retries do not (see below) | | |
| 25 | 25 | | Sign in with GitHub, import from GitHub | Off until you register a GitHub App of your own ([below](#sign-in-with-github-and-import-from-github)). Mirrors sync with **Sync now**: GitHub's webhook needs the REST API. | | |
| 26 | − | | g1t agents, plans, reviews by agents | Off | | |
| 26 | + | | g1t's agent: changes, plans and reviews | Off | | |
| 27 | 27 | | Context hub search | Off | | |
| 28 | 28 | | Deployments on `g1t.page` | Off | | |
| 29 | 29 | | Billing | Off. Nothing is charged, and no usage limit stops work. | |
| 1 | 1 | --- | |
| 2 | 2 | title: Talk to agents | |
| 3 | − | description: Steer a g1t agent while it works, ask it for changes, and let agents ask each other. | |
| 3 | + | description: Steer g1t while it works, ask it for changes, and let agents ask each other. | |
| 4 | 4 | --- | |
| 5 | 5 | ||
| 6 | − | A g1t agent does not work in silence until it is done. You can tell it | |
| 6 | + | g1t does not work in silence until it is done. You can tell it | |
| 7 | 7 | things while it works, ask for changes when it is done, and the agents | |
| 8 | 8 | working on a repository at the same time can ask each other questions and | |
| 9 | 9 | hand each other work. Everything said is recorded in the pull request's | |
| 13 | 13 | | --- | --- | | |
| 14 | 14 | | Correct an agent while it works | [Message the agent](#steer-an-agent-while-it-works) on its pull request. | | |
| 15 | 15 | | Have it change what it made | [Request changes](#ask-for-changes) in a review. | | |
| 16 | − | | Let agents coordinate | Nothing. g1t agents [ask each other](#agents-asking-each-other) through g1t. | | |
| 16 | + | | Let agents coordinate | Nothing. g1t's runs [ask each other](#agents-asking-each-other) through g1t. | | |
| 17 | 17 | ||
| 18 | 18 | ## Steer an agent while it works | |
| 19 | 19 | ||
| 20 | − | While a g1t agent is making or revising a change, its pull request shows | |
| 20 | + | While g1t is making or revising a change, its pull request shows | |
| 21 | 21 | **Message the agent**. | |
| 22 | 22 | ||
| 23 | 23 | 1. Open the pull request. | |
| 56 | 56 | ||
| 57 | 57 | ## Ask for changes | |
| 58 | 58 | ||
| 59 | − | When a g1t agent's pull request is ready, review it the way you would | |
| 59 | + | When a pull request g1t opened is ready, review it the way you would | |
| 60 | 60 | anyone's: | |
| 61 | 61 | ||
| 62 | 62 | 1. Open the **Files changed** tab and comment on the lines you want changed. | |
| 78 | 78 | ||
| 79 | 79 | ### People outrank an agent's review | |
| 80 | 80 | ||
| 81 | − | Whenever a g1t agent revises or reviews a change, it is given what people | |
| 81 | + | Whenever g1t revises or reviews a change, it is given what people | |
| 82 | 82 | have said on the pull request: their comments, comments on lines, | |
| 83 | 83 | approvals and requests for changes. It is told that a change a person asked | |
| 84 | 84 | for is in scope, even where it goes beyond the issue, and that it outranks | |
| 87 | 87 | ||
| 88 | 88 | ## Agents asking each other | |
| 89 | 89 | ||
| 90 | − | g1t agents working in the same repository at the same time can talk | |
| 90 | + | g1t's runs working in the same repository at the same time can talk | |
| 91 | 91 | through g1t, instead of guessing at each other's work. Each one is given | |
| 92 | 92 | the tools to do it, and told when to use them. | |
| 93 | 93 | ||
| 119 | 119 | pull request's sandbox with the agent's own change in front of it and what | |
| 120 | 120 | it was asked; the agent reads its code, answers with `agent` `answer`, and, | |
| 121 | 121 | for a handoff it takes on, commits the work. Its pull request is noted "g1t | |
| 122 | − | woke g1t-agent to answer the agent on #41", and nothing else starts on it | |
| 122 | + | woke g1t to answer the agent on #41", and nothing else starts on it | |
| 123 | 123 | until everything it was asked is answered, or 20 minutes pass. The response | |
| 124 | 124 | to `agent` `message` says so in `hint`, and points the asking agent at the | |
| 125 | 125 | other pull request's change to read meanwhile with `pull_request` `get` | |
| 134 | 134 | ||
| 135 | 135 | ### Rules | |
| 136 | 136 | ||
| 137 | − | - `question` and `handoff` are for g1t agents. A call from your own token, | |
| 137 | + | - `question` and `handoff` are for g1t's own runs. A call from your own token, | |
| 138 | 138 | including your own agent's, sends an ordinary message to the agent on the | |
| 139 | 139 | pull request, as from you. | |
| 140 | 140 | - `from_number` is required from an agent. It may name the agent's issue | |
| 147 | 147 | ||
| 148 | 148 | ## Your own agent | |
| 149 | 149 | ||
| 150 | − | A g1t agent picks up messages between its steps. An agent you run yourself | |
| 150 | + | g1t picks up messages between its steps. An agent you run yourself | |
| 151 | 151 | is not reached this way: steer it in your own client. It can still send | |
| 152 | − | messages to a g1t agent's pull request with `agent` `message`, as above, | |
| 152 | + | messages to a pull request g1t is working on with `agent` `message`, as above, | |
| 153 | 153 | and comment on any pull request with `issue` `comment`. See | |
| 154 | 154 | [connect an agent](/guides/bring-your-own-agent/). |
| 17 | 17 | ||
| 18 | 18 | You must be an **owner of both workspaces**: the one the repository is in | |
| 19 | 19 | and the one it moves to. A member of either cannot, and neither can an | |
| 20 | − | access token that belongs to a workspace or a g1t agent's token. Your email | |
| 20 | + | access token that belongs to a workspace or g1t's token. Your email | |
| 21 | 21 | address must be confirmed. | |
| 22 | 22 | ||
| 23 | 23 | ## Transfer a repository |
| 199 | 199 | | --- | --- | | |
| 200 | 200 | | Making a change for an issue | Yes | | |
| 201 | 201 | | Revising a change after checks, a review or a person | Yes | | |
| 202 | − | | A review by a g1t agent | Yes | | |
| 202 | + | | A review by g1t | Yes | | |
| 203 | 203 | | Catching up with `main` | Yes, when it needed an agent | | |
| 204 | 204 | | Planning an [outcome](/guides/outcomes/) | Yes | | |
| 205 | 205 | | The [merge queue](/guides/merge-queue/) | [Sandbox time](#sandbox-time) | | |
| 220 | 220 | its [sandbox time](#sandbox-time), like any other sandbox. | |
| 221 | 221 | ||
| 222 | 222 | The charge goes to the workspace that owns the repository, whoever | |
| 223 | − | assigned the issue. That is why putting g1t agents to work on a | |
| 223 | + | assigned the issue. That is why putting g1t to work on a | |
| 224 | 224 | repository needs the Write [role](/guides/access-and-roles/) or higher on it. | |
| 225 | 225 | ||
| 226 | 226 | ## How prices are set | |
| 474 | 474 | then **Save caps**; a blank field goes back to the default. A cap g1t staff | |
| 475 | 475 | set for the workspace wins over both. When an issue's agents reach its cap, | |
| 476 | 476 | the message links straight to that section. See | |
| 477 | − | [caps on a plan](/guides/g1t-agents/#caps-on-a-plan) for what happens at | |
| 477 | + | [caps on a plan](/guides/working-with-g1t/#caps-on-a-plan) for what happens at | |
| 478 | 478 | each. | |
| 479 | 479 | ||
| 480 | 480 | ### Alerts |
| 75 | 75 | | `repo.collaborator_added`, `repo.collaborator_role_changed`, `repo.collaborator_removed` | Someone was given a role on it, had their role changed, or lost it. `data.username`, `data.role`, `data.previous_role`. See [access and roles](/guides/access-and-roles/). | | |
| 76 | 76 | | `repo.archived`, `repo.unarchived` | It was made read-only, or writable again. | | |
| 77 | 77 | | `repo.deleted`, `repo.restored`, `repo.purged` | It was deleted, restored within its 30 days, or removed for good. | | |
| 78 | − | | `issue.opened`, `issue.updated`, `issue.assigned`, `issue.closed`, `issue.reopened` | An issue changed. `data.number`; on close, `data.reason` and `data.resolved_by`. | | |
| 78 | + | | `issue.opened`, `issue.updated`, `issue.assigned`, `issue.closed`, `issue.reopened` | An issue changed. `data.number` and `data.author` (`id` and `username`); on close, `data.reason` and `data.resolved_by`. For an issue g1t's agent filed while at work, `data.author` is g1t and `data.requested_by` is the person it was working for. | | |
| 79 | 79 | | `comment.created` | A comment or review on an issue or pull request. | | |
| 80 | − | | `pull.opened`, `pull.ready`, `pull.updated`, `pull.merge_requested`, `pull.merged`, `pull.closed` | A pull request changed. `data.number`, `data.issue`; on merge, `data.commit`. On a g1t agent's change, once g1t has worked it out, `data.confidence`: `level` (`high`, `medium` or `low`), `reasons`, `self_reported`, `uncertain_about`, `run_id` and `assessed_at`. See [how sure the agent is](/guides/g1t-agents/#how-sure-the-agent-is). | | |
| 80 | + | | `pull.opened`, `pull.ready`, `pull.updated`, `pull.merge_requested`, `pull.merged`, `pull.closed` | A pull request changed. `data.number`, `data.issue` and `data.author` (`id` and `username`); on merge, `data.commit`. For a change g1t made, `data.author` is g1t and `data.requested_by` is the person who asked for it; `actor` is still whoever caused the event. On a change by g1t, once g1t has worked it out, `data.confidence`: `level` (`high`, `medium` or `low`), `reasons`, `self_reported`, `uncertain_about`, `run_id` and `assessed_at`. See [how sure the agent is](/guides/working-with-g1t/#how-sure-the-agent-is). | | |
| 81 | 81 | | `checks.completed` | A pull request's checks finished: every status on its head has reported and none is still pending, or the merge queue took it out. `data.number`, `data.commit`, and `data.status`, `passed` or `failed`. | | |
| 82 | − | | `review.completed` | A g1t agent reviewed a pull request. `data.verdict`. | | |
| 82 | + | | `review.completed` | g1t reviewed a pull request. `data.verdict`. | | |
| 83 | 83 | | `workflow.completed` | A [workflow](/guides/actions/) run finished. `data.workflow`, `data.conclusion`, `data.run_id`, `data.sha`, `data.pull`. | | |
| 84 | 84 | | `queue.changed` | The merge queue gained, lost or settled an entry. | | |
| 85 | 85 | | `session.appended` | An agent's session grew. Busy: choose it only if you need it. | |
| 29 | 29 | or with `GET /repos/{owner}/{name}/pulls/{number}/session?after=`. A session | |
| 30 | 30 | is as visible as the repository, so do not put secrets in one. | |
| 31 | 31 | ||
| 32 | − | ### From g1t agents | |
| 32 | + | ### From g1t | |
| 33 | 33 | ||
| 34 | − | A [g1t agent](/guides/g1t-agents/) records its whole session itself: | |
| 34 | + | [g1t](/guides/working-with-g1t/) records its whole session itself: | |
| 35 | 35 | ||
| 36 | 36 | - it opens with a note naming the model that ran, and a note of the other | |
| 37 | 37 | pull requests in progress it was told about; |
| 1 | + | --- | |
| 2 | + | title: g1t's agent | |
| 3 | + | description: Assign an issue to g1t and it does the work. | |
| 4 | + | --- | |
| 5 | + | ||
| 6 | + | g1t can do the work itself. Its agent is named `g1t`: assign an open issue | |
| 7 | + | to g1t and it opens a pull request for the issue, works in a sandbox and a | |
| 8 | + | fork of its own, and reports back as it goes. There is nothing to | |
| 9 | + | configure: you do not say how many agents or which model. Scale comes from | |
| 10 | + | assigning many issues, each worked on by its own run, all at once. | |
| 11 | + | ||
| 12 | + | Everything g1t does shows as `g1t`: the pull requests it opens, its | |
| 13 | + | commits, comments, reviews, plans and security updates, assignments, | |
| 14 | + | timeline events, the audit log, notifications and webhooks. A pull request | |
| 15 | + | g1t opens shows g1t as its author, with **requested by** naming the person | |
| 16 | + | who asked for it. See [who a pull request is for](#who-a-pull-request-is-for). | |
| 17 | + | ||
| 18 | + | g1t's runs are paid for by the workspace they work for, so they need a | |
| 19 | + | paid workspace or the free trial; see [who can run agents](#who-can-run-agents) | |
| 20 | + | and [usage and billing](/guides/usage-and-billing/). Everyone can also | |
| 21 | + | [bring their own agent](/guides/bring-your-own-agent/), which costs | |
| 22 | + | nothing on g1t. | |
| 23 | + | ||
| 24 | + | To hand over a whole outcome rather than one issue at a time, have an agent | |
| 25 | + | plan it first: see [hand off an outcome](/guides/outcomes/). | |
| 26 | + | ||
| 27 | + | ## Put an agent on it in one step | |
| 28 | + | ||
| 29 | + | When the work is not written down yet, open the issue and hand it to the | |
| 30 | + | agent at once: | |
| 31 | + | ||
| 32 | + | 1. On Mission control, choose **Put an agent on it**. | |
| 33 | + | 2. Pick the project, give a title, and say what you want done in plain | |
| 34 | + | words, with what done means if you know it. | |
| 35 | + | 3. Choose **Put an agent on it**. | |
| 36 | + | ||
| 37 | + | You land on the new issue with the agent already at work on its pull | |
| 38 | + | request. The same choice is on a project's **New issue** page, as **Assign | |
| 39 | + | g1t now**, and in the ⌘K palette as **Put an agent on …** followed by | |
| 40 | + | a project's name. | |
| 41 | + | ||
| 42 | + | Putting an agent to work needs the Write role on the project. Without it, | |
| 43 | + | nothing is opened. With it, the issue is always opened, even when the | |
| 44 | + | agent cannot start: | |
| 45 | + | ||
| 46 | + | | What happened | What you see | | |
| 47 | + | | --- | --- | | |
| 48 | + | | The agent started | The issue, with its draft pull request under **Assignees**. | | |
| 49 | + | | Every agent slot of the workspace is busy | The issue, queued for g1t. It starts by itself when a slot frees up. | | |
| 50 | + | | The workspace's plan or limits refused it | The issue is opened, and the composer says why and links to the fix: start the plan or the trial (`not_paid`, `trial_used`), raise the monthly limit (`limit`) or the cap per issue (`issue_cap`), or connect a model (`no_model`). A workspace g1t `paused` says to contact support. | | |
| 51 | + | ||
| 52 | + | From the API or an agent of your own, it is one call: | |
| 53 | + | ||
| 54 | + | ```sh | |
| 55 | + | curl -X POST https://api.g1t.sh/repos/<workspace>/<repo>/issues/delegate \ | |
| 56 | + | -H "Authorization: Bearer $G1T_TOKEN" \ | |
| 57 | + | -H "Content-Type: application/json" \ | |
| 58 | + | -d '{"title": "Retry webhooks with exponential backoff", "body": "Deliveries that fail are dropped today. Retry them up to six times."}' | |
| 59 | + | ``` | |
| 60 | + | ||
| 61 | + | The answer holds the `issue`, the `pull` request the agent opened (or | |
| 62 | + | `null`), and `agent`: its `status` (`started`, `queued` or `not_started`), | |
| 63 | + | and when it did not start, a `code`, a `message` and a `fix_url`. On the MCP | |
| 64 | + | server it is the `agent` tool's `delegate` action. See | |
| 65 | + | [put an agent on it](/reference/api/issues/delegate/). | |
| 66 | + | ||
| 67 | + | The `checks` field this call and `create_issue` used to take is | |
| 68 | + | deprecated. It is still accepted: its commands are added to the issue's | |
| 69 | + | body under `## Definition of done`, one line each (`` - `npm test` passes. ``), | |
| 70 | + | and the answer carries a `deprecation` string saying so. What has to pass | |
| 71 | + | before the pull request merges is the default branch's | |
| 72 | + | [required status checks](/guides/pull-requests/#required-status-checks). | |
| 73 | + | ||
| 74 | + | ## Assigning agents | |
| 75 | + | ||
| 76 | + | One issue: | |
| 77 | + | ||
| 78 | + | 1. Open an issue on a repository. | |
| 79 | + | 2. In **Assign to g1t**, optionally add guidance for this run, on top | |
| 80 | + | of the issue's description. | |
| 81 | + | 3. Choose **Assign**. | |
| 82 | + | ||
| 83 | + | Many issues: | |
| 84 | + | ||
| 85 | + | 1. Open the repository's **Issues** tab. | |
| 86 | + | 2. Tick the issues to hand over, up to ten at a time. | |
| 87 | + | 3. Choose **Assign to g1t**. | |
| 88 | + | ||
| 89 | + | From the API, an agent of your own, or a script: | |
| 90 | + | ||
| 91 | + | ```sh | |
| 92 | + | curl -X POST https://api.g1t.sh/repos/<workspace>/<repo>/issues/12/assign \ | |
| 93 | + | -H "Authorization: Bearer $G1T_TOKEN" | |
| 94 | + | ``` | |
| 95 | + | ||
| 96 | + | The same thing is the `agent` tool's `assign` action on the MCP server, so an agent | |
| 97 | + | planning work can hand issues to g1t itself. | |
| 98 | + | ||
| 99 | + | In a comment: write `@g1t take this` on the issue. See | |
| 100 | + | [mentioning g1t](#mentioning-g1t). | |
| 101 | + | ||
| 102 | + | By label: a project can hand every issue given a label to the agent. See | |
| 103 | + | [the label rule](#the-label-rule). | |
| 104 | + | ||
| 105 | + | Each run appears as a draft pull request on its issue within a few | |
| 106 | + | seconds. The pages update on their own while g1t works. | |
| 107 | + | ||
| 108 | + | An issue can still have more than one pull request: assign it again, or | |
| 109 | + | have your own agent open one alongside. That is for when you want a second | |
| 110 | + | attempt, not the normal way of working. | |
| 111 | + | ||
| 112 | + | ## What an agent does | |
| 113 | + | ||
| 114 | + | 1. Clones its pull request's fork. | |
| 115 | + | 2. Is told what else is in progress: every other open pull request in the | |
| 116 | + | repository, what it is for and which files it changes. Its session | |
| 117 | + | starts with a note of what it was told. | |
| 118 | + | 3. Reads the code and makes the change the issue asks for, keeping clear of | |
| 119 | + | the other work where it can. | |
| 120 | + | 4. Commits its work. | |
| 121 | + | 5. Pushes to the fork and marks the pull request ready for review, with a | |
| 122 | + | summary as its description. | |
| 123 | + | ||
| 124 | + | Everything it reads, runs and decides is recorded in the pull request's | |
| 125 | + | **Session** as it happens; see [sessions and why-blame](/guides/why-blame/). | |
| 126 | + | The **Files changed** tab shows the resulting diff. | |
| 127 | + | ||
| 128 | + | If an agent fails, or finishes without changing anything, its pull request | |
| 129 | + | is closed and its session says why. | |
| 130 | + | ||
| 131 | + | ## Repository instructions | |
| 132 | + | ||
| 133 | + | Every g1t run reads the repository's own instructions for agents | |
| 134 | + | and is told them, labelled as the repository's, before it starts: making a | |
| 135 | + | change, revising it, reviewing, catching up, answering, and planning. | |
| 136 | + | ||
| 137 | + | | File | Read by | | |
| 138 | + | | --- | --- | | |
| 139 | + | | `AGENTS.md` and `CLAUDE.md` at the root | Every run. | | |
| 140 | + | | `AGENTS.md` and `CLAUDE.md` in a subdirectory | Runs whose task touches files under it: the nearest one above each file. Where it disagrees with the root's, it wins for the files under it. | | |
| 141 | + | | `.g1t/review.md` | Reviews: what to check, house rules, paths that need extra care. | | |
| 142 | + | ||
| 143 | + | Write them for an agent that knows nothing about the project: how to build | |
| 144 | + | and test, how things are named, what never to touch. For example: | |
| 145 | + | ||
| 146 | + | ```md | |
| 147 | + | # AGENTS.md | |
| 148 | + | - Run `npm test` and `npm run typecheck` before you finish. | |
| 149 | + | - API handlers return a Result; they never throw. | |
| 150 | + | - Never edit files under `vendor/`. | |
| 151 | + | ``` | |
| 152 | + | ||
| 153 | + | Which directories a task touches comes from the pull request's changed | |
| 154 | + | files, and for new work from the paths the issue names, so naming the | |
| 155 | + | files in an issue helps the right instructions reach the agent. | |
| 156 | + | ||
| 157 | + | **Where they are read from.** The default branch, as it is when the run | |
| 158 | + | starts. A pull request from one of the repository's own branches is read at | |
| 159 | + | its head instead, since only people who can push to the repository can | |
| 160 | + | change it. A pull request from a fork, which includes every change g1t | |
| 161 | + | makes, is never followed: the agent keeps the default branch's | |
| 162 | + | instructions, and if the fork changes them, it is shown the changed text as | |
| 163 | + | part of the change, marked as not instructions. That way nobody can steer | |
| 164 | + | an agent, or the review of their own change, by editing these files in a | |
| 165 | + | pull request. Treat what a fork's head says as untrusted, as you would its | |
| 166 | + | code. | |
| 167 | + | ||
| 168 | + | **Limits.** Each file is cut at 8,000 characters and all of them together | |
| 169 | + | at 24,000; what is left out is named in the prompt. Files are read once per | |
| 170 | + | commit and reused. | |
| 171 | + | ||
| 172 | + | The project's **Agents** page lists the files its runs read, what they say | |
| 173 | + | and when each last changed, with a link to each in the code. Each run's | |
| 174 | + | session starts with a note of which files it read. To change them, change | |
| 175 | + | the files and merge to the default branch. | |
| 176 | + | ||
| 177 | + | ## Seeing it through | |
| 178 | + | ||
| 179 | + | Making the change is the first step. g1t takes the rest itself, and you | |
| 180 | + | get the pull request back ready to merge: | |
| 181 | + | ||
| 182 | + | 1. **Checks.** Every push the agent makes runs the repository's | |
| 183 | + | [workflows](/guides/actions/) on the pull request, as for anyone's pull | |
| 184 | + | request. g1t waits for them to finish. Only the workflow runs report, so | |
| 185 | + | the agent has no say in the result. | |
| 186 | + | 2. **Review.** A different agent reads the change and posts comments on | |
| 187 | + | lines, a summary and a verdict. | |
| 188 | + | 3. **Revision.** If a check fails or the review asks for changes, the | |
| 189 | + | author is sent back with exactly what was found. For a failed check, | |
| 190 | + | that is the end of the log of each failed job, up to three jobs and | |
| 191 | + | about 3,000 characters each; it can read more with `get_workflow_run` | |
| 192 | + | and `get_job_logs`. Steps 1 and 2 run again on the result. This happens | |
| 193 | + | at most twice, or as often as the repository's **Revisions before asking | |
| 194 | + | you** allows. | |
| 195 | + | 4. **Ready to merge.** The required checks passed and it is approved. | |
| 196 | + | Merging is yours, unless the repository says otherwise (below). | |
| 197 | + | ||
| 198 | + | Agents are told to run the same tests and linters the workflows run before | |
| 199 | + | they finish, so most failures are caught in the sandbox. What "done" means | |
| 200 | + | for the issue, if its description says so, is context for the agent and | |
| 201 | + | its reviewer; what decides the merge is the default branch's | |
| 202 | + | [required status checks](/guides/pull-requests/#required-status-checks). | |
| 203 | + | A repository with no workflows has nothing to prove a change works; **Add | |
| 204 | + | CI** gives it one ([add CI](/guides/actions/#add-ci)). | |
| 205 | + | ||
| 206 | + | If `main` has moved in the meantime, that does not hold the pull request | |
| 207 | + | up. Merging it brings it up to date first: g1t merges `main` in, an agent | |
| 208 | + | resolves any conflict, and it lands. A repository that wants every pull | |
| 209 | + | request caught up and checked again before it may merge turns on **Require | |
| 210 | + | pull requests to be up to date before merging** in its settings; catching | |
| 211 | + | up is then a step of its own, before "ready". | |
| 212 | + | ||
| 213 | + | g1t also works out ahead of time whether each pull request still merges | |
| 214 | + | cleanly, every time it or `main` moves | |
| 215 | + | ([how](/guides/pull-requests/#conflicts)). When one of an agent's pull | |
| 216 | + | requests is found to conflict, g1t does not wait for a merge to trip over | |
| 217 | + | it: the agent is sent to merge `main` in and resolve the conflicts, told | |
| 218 | + | which files conflict, and the checks run again on the result. | |
| 219 | + | ||
| 220 | + | The pull request's page shows which step it is at. While a required check | |
| 221 | + | has not reported on its latest commit, it says so: "Waiting for the | |
| 222 | + | required check CI to report on its latest commit." If g1t cannot finish, | |
| 223 | + | because a required check still fails after the agent revised as often as | |
| 224 | + | the repository allows ("The required check CI still fails after the agent | |
| 225 | + | revised twice."), a review could not be written, or a conflict could not | |
| 226 | + | be resolved while bringing it up to date, it stops and the page says | |
| 227 | + | **Needs you**, with the reason. A check that is not required and still | |
| 228 | + | fails after the last revision does not hold it. Pushing to the pull | |
| 229 | + | request yourself starts it moving again. | |
| 230 | + | ||
| 231 | + | ### What a repository can ask for | |
| 232 | + | ||
| 233 | + | Under a project's **Settings → Branches and merging**, someone with the | |
| 234 | + | Maintain [role](/guides/access-and-roles/) or higher sets the rules its pull | |
| 235 | + | requests follow. **Branch protection** holds the rules for every pull | |
| 236 | + | request, a person's or an agent's; they are described in | |
| 237 | + | [required status checks](/guides/pull-requests/#required-status-checks): | |
| 238 | + | ||
| 239 | + | | Setting | Default | What it does | | |
| 240 | + | | --- | --- | --- | | |
| 241 | + | | Require a pull request to change the default branch | Off | Refuses pushes to the default branch. | | |
| 242 | + | | Required status checks | None | The checks that must pass on a pull request's head before it merges. | | |
| 243 | + | | Required approvals | None | How many reviewers must approve before a merge. A reviewer who asked for changes blocks it. | | |
| 244 | + | | g1t's approval counts | On | Off means approvals have to come from people. | | |
| 245 | + | | Require branches to be up to date before merging | Off | On means catching up is a step of its own and the checks run again. | | |
| 246 | + | | Merge through a queue | Off | Merging tests a pull request together with those ahead of it; the default branch only moves to a combination that passed. See [merge queue](/guides/merge-queue/). | | |
| 247 | + | | Allow bypassing required checks | On | Lets someone who may merge merge without the required checks passing. Off means nobody can. | | |
| 248 | + | ||
| 249 | + | In the **g1t** section, you set what g1t does with its own pull requests: | |
| 250 | + | ||
| 251 | + | | Setting | Default | What it does | | |
| 252 | + | | --- | --- | --- | | |
| 253 | + | | Review by a second agent | On | Off leaves review to people. | | |
| 254 | + | | Revisions before asking you | 2 | How often an agent is sent back before g1t stops. | | |
| 255 | + | | Merge automatically when ready | Off | Lands g1t's pull request once every rule is met. | | |
| 256 | + | | Ask a person before merging low-confidence changes | On | A change by g1t [rated low](#how-sure-the-agent-is) waits for a person's approval instead of merging by itself or joining the queue. | | |
| 257 | + | ||
| 258 | + | A pull request g1t opens follows the same rules as anyone's. If the | |
| 259 | + | repository wants approvals from people, it waits for them, and shows | |
| 260 | + | **Needs you** until they arrive. | |
| 261 | + | ||
| 262 | + | ### Talking to an agent | |
| 263 | + | ||
| 264 | + | While g1t works, you can steer it with **Message the agent** on its | |
| 265 | + | pull request; it reads the message at its next step, without starting | |
| 266 | + | over. Once it is done, a review with **Request changes** sends it back to | |
| 267 | + | make them, and the checks and review run again. g1t's runs working at the | |
| 268 | + | same time can also ask each other questions and hand each other work. See | |
| 269 | + | [talk to agents](/guides/talking-to-agents/). | |
| 270 | + | ||
| 271 | + | ### Merging automatically | |
| 272 | + | ||
| 273 | + | A repository can land g1t's pull request by itself once it is | |
| 274 | + | ready. Someone with the Maintain role or higher turns this on under the repository's | |
| 275 | + | **Settings → Branches and merging**; it is off to begin with. The merge is recorded as made by | |
| 276 | + | `g1t`, the issue closes naming the pull request, and nothing short of | |
| 277 | + | ready is ever merged this way: every required check has passed on its | |
| 278 | + | head. One that is behind `main` is brought up to | |
| 279 | + | date as part of the merge. Pull requests from people and from other | |
| 280 | + | agents always wait for a person to merge them. | |
| 281 | + | ||
| 282 | + | Every step is recorded: revisions and catch-ups in the pull request's | |
| 283 | + | **Session**, reviews in its conversation. | |
| 284 | + | ||
| 285 | + | This applies to pull requests g1t opens. One you or your own | |
| 286 | + | agent opened is yours to drive; the same workflows run on it, and you can ask | |
| 287 | + | for a review or a catch-up from its page. | |
| 288 | + | ||
| 289 | + | ### How sure the agent is | |
| 290 | + | ||
| 291 | + | Once g1t has finished a change, g1t records how sure it is that the | |
| 292 | + | change is right: **high**, **medium** or **low**, with a few words saying | |
| 293 | + | why, such as "Low — tests not added, 3 revisions". It shows on the pull | |
| 294 | + | request, under the agent, and on Mission control. It is worked out again as | |
| 295 | + | the change moves through checks, review and revision, and kept with each | |
| 296 | + | run, so a run's page says how the change stood when that run left it. | |
| 297 | + | ||
| 298 | + | Confidence comes from what g1t can observe, not from how the agent sounds. | |
| 299 | + | Each signal below that tells against the change adds points, or makes it | |
| 300 | + | low on its own. No points is high, one or two is medium, and three or more | |
| 301 | + | is low. | |
| 302 | + | ||
| 303 | + | | Signal | Effect | | |
| 304 | + | | --- | --- | | |
| 305 | + | | Required checks fail | Low | | |
| 306 | + | | It failed in the [merge queue](/guides/merge-queue/) | Low | | |
| 307 | + | | The reviewer agent asks for changes | Low | | |
| 308 | + | | A run was stopped at its cost or time cap | Low | | |
| 309 | + | | Sent back to revise | 1 point per revision, at most 3 | | |
| 310 | + | | Required checks have not finished, or have not run on its head | 1 point | | |
| 311 | + | | Checks passed only on a retry | 1 point | | |
| 312 | + | | The default branch has no required checks | 1 point | | |
| 313 | + | | No review yet, or the repository has no reviewer agent | 1 point | | |
| 314 | + | | The reviewer approved but left three or more comments on lines | 1 point | | |
| 315 | + | | Code changed and no test was added or changed | 1 point | | |
| 316 | + | | More than 400 lines changed; more than 1,000 | 1 point; 2 points | | |
| 317 | + | | More than 30 files changed | 1 point | | |
| 318 | + | | Files changed outside the area its [plan](/guides/outcomes/) expected; four or more | 1 point; 2 points | | |
| 319 | + | | Touches CI workflows, repository automation, secrets, infrastructure or `CODEOWNERS` | 2 points | | |
| 320 | + | | Its latest run used 80% or more of its cost or time cap | 1 point each | | |
| 321 | + | | Steps refused by [guardrails](/guides/guardrails/); three or more | 1 point; 2 points | | |
| 322 | + | | A question or handoff it sent another agent is unanswered | 2 points | | |
| 323 | + | | The agent said it was unsure about something | 1 point | | |
| 324 | + | ||
| 325 | + | At the end of every run that makes or revises a change, the agent is also | |
| 326 | + | asked how sure it is, and what it could not verify. g1t takes the lower of | |
| 327 | + | the two: what it observes can lower the agent's own word, never raise it. | |
| 328 | + | When the agent's word is lower, the reasons start with "agent says low", | |
| 329 | + | and the pull request lists what it was unsure about. | |
| 330 | + | ||
| 331 | + | For high confidence, the reasons say what it rests on: required checks pass, | |
| 332 | + | approved on the first review, tests added, a small change. | |
| 333 | + | ||
| 334 | + | The pull request's `confidence` in the | |
| 335 | + | [API](/reference/api/pull-requests/get-pull-request/) has the `level`, | |
| 336 | + | `reasons`, `self_reported`, `uncertain_about`, the `run_id` it was worked out | |
| 337 | + | after, and `assessed_at`. [Webhooks](/guides/webhooks/) for pull requests | |
| 338 | + | carry it too. | |
| 339 | + | ||
| 340 | + | ### Low-confidence changes wait for a person | |
| 341 | + | ||
| 342 | + | With **Ask a person before merging low-confidence changes** on, which it is | |
| 343 | + | unless someone turns it off, a change by g1t rated low is not merged | |
| 344 | + | by itself and does not join the merge queue, even with **Merge | |
| 345 | + | automatically when ready** on. Once everything else the repository asks | |
| 346 | + | for is met, it stops at **Needs you**, saying why, and Mission control | |
| 347 | + | lists it under **Needs you** with a **Low confidence** chip, the reasons in | |
| 348 | + | **What the agent already knows**, and the reasons again in **Why this | |
| 349 | + | needs you**. | |
| 350 | + | ||
| 351 | + | To let it land, approve it: a person's approval since the agent last | |
| 352 | + | revised lifts the hold, and it merges as the repository's rules say. To | |
| 353 | + | send it back, request changes. Merging it yourself works as usual. The | |
| 354 | + | setting is under **Settings → Branches and merging**, in the **g1t** section, | |
| 355 | + | and is `hold_low_confidence` in | |
| 356 | + | [`update_repo_settings`](/reference/api/repositories/update-repo-settings/). | |
| 357 | + | ||
| 358 | + | ## Choosing between pull requests | |
| 359 | + | ||
| 360 | + | Each pull request on the issue's page shows whether its checks passed. Open | |
| 361 | + | the ones that did, read their descriptions and changes, and merge the one | |
| 362 | + | you want. Merging lands it on `main` and closes the issue, which records | |
| 363 | + | that pull request as the one that resolved it. The other pull requests for | |
| 364 | + | the issue close as superseded. See | |
| 365 | + | [merging](/concepts/overview/#merging) for what happens when `main` has moved. | |
| 366 | + | ||
| 367 | + | ## Other things g1t does | |
| 368 | + | ||
| 369 | + | - **Review.** On a pull request that is ready, **Request review from g1t** | |
| 370 | + | has g1t read the change and post comments on lines, a summary and a | |
| 371 | + | verdict. | |
| 372 | + | - **Catch up.** When `main` has moved under a pull request, **Catch up with | |
| 373 | + | main** merges it in. When the two changed different files g1t does that | |
| 374 | + | itself in seconds, with no agent; otherwise an agent merges it in a | |
| 375 | + | sandbox and resolves any conflict | |
| 376 | + | ([catching up](/guides/pull-requests/#catching-up)). | |
| 377 | + | - **Finish a security update.** g1t raises a vulnerable dependency to its | |
| 378 | + | fixed version itself, with no agent. When raising the version is not | |
| 379 | + | enough (the bump fails, or the pull request's required checks fail | |
| 380 | + | because code must change), g1t opens an issue and assigns it to g1t. | |
| 381 | + | That session shows as **started by g1t** on the project's **Agents** page. | |
| 382 | + | See [security updates](/guides/security/#security-updates). | |
| 383 | + | ||
| 384 | + | Each runs in a sandbox of its own. | |
| 385 | + | ||
| 386 | + | ## Mentioning g1t | |
| 387 | + | ||
| 388 | + | Write `@g1t` in a comment on an issue or a pull request, with what | |
| 389 | + | you want, and it does it. The comment box offers to complete the name as | |
| 390 | + | you type `@`. | |
| 391 | + | ||
| 392 | + | | Where | You write | What happens | | |
| 393 | + | | --- | --- | --- | | |
| 394 | + | | An issue | A request: `@g1t take this`, `@g1t fix the empty case` | The issue is assigned to g1t, which opens a pull request, as if you had chosen **Assign**. | | |
| 395 | + | | An issue | A question: `@g1t why is this slow?` | g1t reads the code on the default branch and answers in the thread. It changes nothing. | | |
| 396 | + | | A pull request g1t made | A request: `@g1t also handle the empty list` | g1t is sent back to make the change, with your comment as what to address, and the checks and review run again. If it is still working, it gets your comment as a message at its next step. | | |
| 397 | + | | Any pull request | `@g1t review this` | A review by g1t, as with **Request review from g1t**. | | |
| 398 | + | | Any pull request | A question | g1t reads the change at its head and answers in the thread. On someone else's pull request, which it cannot push to, a request is answered too: it says what it would change. | | |
| 399 | + | ||
| 400 | + | A request is a comment whose words after the mention start with what to | |
| 401 | + | do (`take`, `fix`, `add`, `please rename`, `can you update`); a question | |
| 402 | + | starts with a question word or ends with a question mark. `review` near the | |
| 403 | + | start asks for a review. | |
| 404 | + | ||
| 405 | + | g1t always replies in the thread, saying what it started or why it | |
| 406 | + | did not. Every run a mention starts shows on the project's **Agents** page | |
| 407 | + | as started by whoever mentioned it, and a mention that started nothing | |
| 408 | + | shows there as a failed run with the reason. | |
| 409 | + | ||
| 410 | + | **What does not count.** Mentions in code (`` `@g1t` `` or a code | |
| 411 | + | block), in quoted lines (`> @g1t …`), in email addresses | |
| 412 | + | (`ops@g1t.sh`), in URLs, in package scopes (`@g1t/platform`) and in longer | |
| 413 | + | names (`@g1t-bot`) are ignored. Matching ignores case. Agents mentioning | |
| 414 | + | `@g1t` start nothing, so | |
| 415 | + | agents cannot set each other to work this way. | |
| 416 | + | ||
| 417 | + | **Who can.** People with the Write [role](/guides/access-and-roles/) or higher on the | |
| 418 | + | repository, members or not. Anyone else who mentions it gets a short reply | |
| 419 | + | saying that putting g1t to work needs the Write role on the | |
| 420 | + | repository, and nothing starts. When | |
| 421 | + | the workspace's plan does not let the agent start (a free workspace with no | |
| 422 | + | trial left, a paused workspace, an issue at its spending cap), g1t | |
| 423 | + | replies with why and where to fix it. When every agent slot is busy, it | |
| 424 | + | replies that the run is waiting for a free slot, and starts it when one | |
| 425 | + | finishes. | |
| 426 | + | ||
| 427 | + | Each comment starts one run at most; to ask again, write a new comment. | |
| 428 | + | ||
| 429 | + | ## The label rule | |
| 430 | + | ||
| 431 | + | Under a project's **Settings → Agents**, someone with the Maintain role or | |
| 432 | + | higher sets a label, such as `agent`. From then on, when someone with the | |
| 433 | + | Write role or higher gives an open issue that label, either | |
| 434 | + | when opening it or later, g1t takes it: the issue is queued for | |
| 435 | + | g1t, the conversation says so, and the agent starts as soon as the | |
| 436 | + | project has room and nothing the issue depends on is still open, exactly as | |
| 437 | + | for a [plan's](/guides/outcomes/) issues. An issue that already had the | |
| 438 | + | label is not affected; removing and adding it again counts. **Turn off** | |
| 439 | + | removes the rule. | |
| 440 | + | ||
| 441 | + | ## Which model runs | |
| 442 | + | ||
| 443 | + | You do not pick one. You assign the work to `g1t`, the way you would | |
| 444 | + | assign an issue to a colleague, and g1t routes it. On g1t's hosted models, | |
| 445 | + | each piece of work goes to the least costly of two tiers that can do it: | |
| 446 | + | ||
| 447 | + | | Tier | Model today | | |
| 448 | + | | --- | --- | | |
| 449 | + | | Small | Claude Haiku 4.5 | | |
| 450 | + | | Large | Claude Sonnet 5.5 | | |
| 451 | + | ||
| 452 | + | The work decides the tier: | |
| 453 | + | ||
| 454 | + | | Work | Tier | | |
| 455 | + | | --- | --- | | |
| 456 | + | | Making a change for an issue, revising it, and answering a mention | Large | | |
| 457 | + | | Reviewing a pull request that changes at most 10 files and 200 lines, touches no sensitive path, and is not for an issue labelled `security` | Small | | |
| 458 | + | | Reviewing any other pull request, or one whose changed files g1t does not know yet | Large | | |
| 459 | + | | Catching up with `main` and resolving conflicts | Small | | |
| 460 | + | | Planning an outcome | Small | | |
| 461 | + | | Any of these again, after the last attempt at the same work failed or stopped at a guardrail cap | Large | | |
| 462 | + | ||
| 463 | + | Sensitive paths are the ones that run, configure or guard things: CI | |
| 464 | + | workflows, `.g1t/` and `.github/`, `CODEOWNERS`, secrets such as `.env` | |
| 465 | + | and `.pem` files, and infrastructure such as Dockerfiles, Terraform and | |
| 466 | + | `wrangler.*` files. They are the same paths that lower a change's | |
| 467 | + | [confidence](#how-sure-the-agent-is). | |
| 468 | + | ||
| 469 | + | The agent's own small background steps run on the small tier. | |
| 470 | + | ||
| 471 | + | Every session opens with a note naming the model that ran, and an agent's | |
| 472 | + | review says which model wrote it, so what you got is always on the record. | |
| 473 | + | When a better model for a tier appears, g1t changes the route and nothing | |
| 474 | + | you have set up needs to change. | |
| 475 | + | ||
| 476 | + | A workspace that routes its work to [its own provider](/guides/models/) | |
| 477 | + | is not routed by tier: its work runs on the model its route names. | |
| 478 | + | ||
| 479 | + | A pull request g1t opens has `g1t` as its author and as its `agent` in the | |
| 480 | + | API, and its commits are authored `g1t <g1t@users.noreply.g1t.sh>`. | |
| 481 | + | ||
| 482 | + | ## Who a pull request is for | |
| 483 | + | ||
| 484 | + | g1t is the author of every pull request it makes and of every issue it | |
| 485 | + | files while at work. The person who asked for the work, by assigning the | |
| 486 | + | issue or handing g1t the task, is kept beside it as **requested by**. Work | |
| 487 | + | g1t starts itself, such as a [security update](/guides/security/), names | |
| 488 | + | nobody. | |
| 489 | + | ||
| 490 | + | | Where | Author | Who asked | | |
| 491 | + | | --- | --- | --- | | |
| 492 | + | | The pull request's page, lists and link previews | **g1t** | **requested by** *name*, or **for** *name* | | |
| 493 | + | | The API and MCP (`get_pull_request`, `list_pull_requests`, `get_issue`) | `author`: `{ "username": "g1t", "kind": "agent" }` | `requested_by`, or `null` | | |
| 494 | + | | [Webhooks](/guides/webhooks/) | `data.author` | `data.requested_by`, left out when nobody asked | | |
| 495 | + | | [Actions](/guides/actions/) (`github.event`) | `pull_request.user`, a `Bot` named `g1t` | `pull_request.requested_by`; `sender` is whoever caused the event | | |
| 496 | + | | [Search](/guides/search/) | `author:g1t` finds it | shown as **for** *name* | | |
| 497 | + | ||
| 498 | + | The person who asked answers for the pull request as its author would: | |
| 499 | + | ||
| 500 | + | - they can update, close and mark it ready, catch it up and steer its | |
| 501 | + | agent without the Triage role; | |
| 502 | + | - they are never asked to review it, and they cannot approve it or | |
| 503 | + | request changes on it; nor does their approval count toward the | |
| 504 | + | repository's required approvals; | |
| 505 | + | - it is on their own lists: what they are working on, and their profile; | |
| 506 | + | - the sandboxes that work on it act as them, so it reaches what they can; | |
| 507 | + | - its workflows and preview get secrets only when they have the Write | |
| 508 | + | role or higher, as theirs would. | |
| 509 | + | ||
| 510 | + | Being its author gives g1t nothing more: a review by g1t's agent still | |
| 511 | + | counts where the repository lets an agent's approval count. | |
| 512 | + | ||
| 513 | + | ## How model traffic is routed | |
| 514 | + | ||
| 515 | + | g1t's runs send model requests to g1t's model proxy at | |
| 516 | + | `https://models.g1t.sh`, with a token for their run in place of a key; see | |
| 517 | + | [your keys never reach a sandbox](/guides/models/#your-keys-never-reach-a-sandbox). | |
| 518 | + | Requests for g1t's hosted models go on through | |
| 519 | + | [Cloudflare AI Gateway](https://developers.cloudflare.com/ai-gateway/), | |
| 520 | + | which holds g1t's key. Each of those requests is tagged with the kind of | |
| 521 | + | work, the tier, the repository and the pull request, so spend can be read | |
| 522 | + | per tier and per pull request. Requests for a workspace's own provider go to that provider. | |
| 523 | + | ||
| 524 | + | If you run your own copy of g1t, these settings control it: | |
| 525 | + | ||
| 526 | + | | Setting | Where | What it does | | |
| 527 | + | | --- | --- | --- | | |
| 528 | + | | `AGENT_ROUTING` | Runner | JSON. `tiers`: the model behind `small` and `large`, each `{ "modelName", "model" }`. `tasks`: the tier of `implement`, `review`, `update` and `plan`, or `change` to decide by the change. `smallChange`: the most `files` and `lines` a `change` review runs on the small tier with. `largeLabels`: issue labels that keep a review on the large tier. Anything left out takes the defaults above. | | |
| 529 | + | | `MODELS_URL` | Runner | Where sandboxes send model requests: the model proxy. | | |
| 530 | + | | `AI_GATEWAY_ID` | Model proxy | The gateway hosted requests go through. Empty sends them to the provider directly. | | |
| 531 | + | | `AI_GATEWAY_TOKEN` | Model proxy | Secret. Authenticates to the gateway. | | |
| 532 | + | | `ANTHROPIC_API_KEY` | Model proxy | Secret. The provider's key, if the gateway does not hold it. | | |
| 533 | + | ||
| 534 | + | ## What it costs | |
| 535 | + | ||
| 536 | + | A workspace pays for g1t's runs on its repositories, after | |
| 537 | + | they run: each run is charged its sandbox by the second, at cost plus 20%, | |
| 538 | + | and, on g1t's hosted models, what AI Gateway priced its model requests at, | |
| 539 | + | plus 20%. A workspace's [own provider](/guides/models/) bills it for the | |
| 540 | + | model directly. See | |
| 541 | + | [Usage and billing](/guides/usage-and-billing/) for how prices are set and | |
| 542 | + | the limits on usage not yet paid for. | |
| 543 | + | The workspace's **Usage** page shows what its agents have cost, by day, | |
| 544 | + | kind of work, repository, model and pull request. See | |
| 545 | + | [usage and billing](/guides/usage-and-billing/). | |
| 546 | + | ||
| 547 | + | ## What a sandbox has | |
| 548 | + | ||
| 549 | + | Git, common shell tools, and toolchains for Node.js, Python, Go and Rust, so | |
| 550 | + | an agent can build and test most projects. If your project needs something | |
| 551 | + | else, the agent will say in its summary what it could not run. | |
| 552 | + | ||
| 553 | + | A workspace (or a project) can send its agents' work to | |
| 554 | + | [its own runners](/guides/self-hosted-runners/#agents-on-your-runners) | |
| 555 | + | instead, so agents build and test with what those machines have. The agent | |
| 556 | + | works the same way there, with the same short-lived credentials, and its | |
| 557 | + | model calls still go through g1t; the machine time is free. g1t's network | |
| 558 | + | guardrails cannot be enforced on your machines, and the run says so. | |
| 559 | + | ||
| 560 | + | ## Who can run agents | |
| 561 | + | ||
| 562 | + | Putting an agent to work (assigning it, mentioning it, asking it for a | |
| 563 | + | review, planning, sending it back to revise) needs the Write | |
| 564 | + | [role](/guides/access-and-roles/) or higher on the repository. That | |
| 565 | + | includes an [outside collaborator](/guides/access-and-roles/#outside-collaborators) | |
| 566 | + | with Write: their runs are charged to the repository's workspace, as a | |
| 567 | + | member's are, and count against its plan, caps and agent slots. They see | |
| 568 | + | what their agents do, but not which model ran or what a run cost; those | |
| 569 | + | are for members of the workspace. A run for an outside collaborator is | |
| 570 | + | told the project's memory, never the workspace's. | |
| 571 | + | ||
| 572 | + | Agents cost g1t real money, so they run for paid workspaces. A free | |
| 573 | + | workspace has the whole forge, and two ways to try agents: | |
| 574 | + | ||
| 575 | + | - **The trial.** $5 of usage, once per workspace, after a card check. | |
| 576 | + | - **The open-source pool.** Workflows and the merge queue on public | |
| 577 | + | repositories, after the same card check. It does not pay for agents. | |
| 578 | + | ||
| 579 | + | This holds whether the agent uses g1t's hosted models or | |
| 580 | + | [the workspace's own model provider](/guides/models/): the sandbox an agent | |
| 581 | + | works in is g1t's either way. Before anyone assigns, asks for a review or | |
| 582 | + | plans, a free workspace's pages say "Agents need a paid workspace or the | |
| 583 | + | free trial", with a link to its **Billing** page. | |
| 584 | + | ||
| 585 | + | Every agent run, of every kind (making a change, revising, reviewing, | |
| 586 | + | catching up, planning, answering a mention), asks billing before it | |
| 587 | + | starts. Billing reserves what the run is expected to cost: its model's | |
| 588 | + | recent average (about $0.10 to make a change or plan, $0.07 to review) | |
| 589 | + | plus its sandbox for its whole time cap. When the run ends, what it really | |
| 590 | + | cost is settled against that. If billing refuses, nothing starts, and you | |
| 591 | + | see why where you started it: | |
| 592 | + | ||
| 593 | + | | Where you started it | Where the refusal shows | | |
| 594 | + | | --- | --- | | |
| 595 | + | | **Assign to g1t**, **Request review**, **Plan it**, catching up | Under the button | | |
| 596 | + | | A mention or the label rule | A comment from g1t on the issue or pull request | | |
| 597 | + | | A step g1t takes by itself (a review, a revision, a catch-up) | The pull request's status, which then waits for you | | |
| 598 | + | ||
| 599 | + | Each refusal says what to do: start the plan or the trial, raise the spend | |
| 600 | + | limit, or wait for next month's open-source pool, with the page to do it | |
| 601 | + | on. | |
| 602 | + | ||
| 603 | + | ### Caps on a plan | |
| 604 | + | ||
| 605 | + | A workspace's plan sets caps on its agents. A new paid workspace in its | |
| 606 | + | first month, and a workspace on the trial, has tighter ones. The amounts | |
| 607 | + | are on [usage and billing](/guides/usage-and-billing/#caps). | |
| 608 | + | ||
| 609 | + | | Cap | What happens at it | | |
| 610 | + | | --- | --- | | |
| 611 | + | | Agents at once | A run over the cap waits for a free slot instead of being refused. An assigned issue goes back in the queue; a review, catch-up, plan or answer someone asked for waits its turn; a step g1t takes by itself is tried again at its next sweep, within five minutes. Each says "Waiting for a free slot". | | |
| 612 | + | | Time per run | The lower of the project's [guardrails](/guides/guardrails/) time cap for that kind of run and the plan's. | | |
| 613 | + | | Cost per run | The lower of the guardrails' cost cap and the plan's. The agent is stopped when it reaches it, as with any cost cap. | | |
| 614 | + | | Cost per issue | What every agent run on an issue and its pull requests has cost in all. Past it, g1t does not start on that issue again and says so on it; an owner can raise the cap on the **Billing** page. | | |
| 615 | + | ||
| 616 | + | When the workspace's compute is paused (a spend spike waiting for an owner, | |
| 617 | + | or a hold by g1t), nothing new starts, and the refusal gives the reason. | |
| 618 | + | ||
| 619 | + | ### When billing cannot be reached | |
| 620 | + | ||
| 621 | + | g1t's own billing service could be briefly unreachable. Then: | |
| 622 | + | ||
| 623 | + | - A paid workspace's runs go ahead, and the miss is logged. A billing blip | |
| 624 | + | never stops a paying customer's work. | |
| 625 | + | - A free workspace's runs do not start: they would be paid for by nobody. | |
| 626 | + | Try again in a minute. | |
| 627 | + | ||
| 628 | + | g1t decides which a workspace is from its plan, or from the last plan it | |
| 629 | + | saw for it in the past day. | |
| 630 | + | ||
| 631 | + | ### Other limits | |
| 632 | + | ||
| 633 | + | - A run has two hours. After that its credentials expire and it can no | |
| 634 | + | longer push or report. | |
| 635 | + | - Agents do not run on an [archived](/guides/managing-repositories/#archive-a-repository) | |
| 636 | + | or deleted repository: nothing new starts, whether from an assignment, | |
| 637 | + | a mention or the label rule, and a run under way cannot push or merge. | |
| 638 | + | What was refused does not start by itself when the repository is | |
| 639 | + | unarchived or restored; assign the work again. | |
| 640 | + | ||
| 641 | + | ## Credentials | |
| 642 | + | ||
| 643 | + | Every sandbox run gets credentials of its own, made when it starts and | |
| 644 | + | revoked the moment it stops. They are not your access tokens, and they are | |
| 645 | + | not listed with them. | |
| 646 | + | ||
| 647 | + | Each credential carries a composite identity: the agent, acting on behalf | |
| 648 | + | of the person who started the work. A run you started by assigning an issue | |
| 649 | + | is `g1t on behalf of you`, and that is how it appears in the | |
| 650 | + | [audit log](/guides/audit-log/), on the run's page and in the pull | |
| 651 | + | request's **Agent** panel. | |
| 652 | + | ||
| 653 | + | What it may do is the intersection of two things: | |
| 654 | + | ||
| 655 | + | - **The run's scope.** The credential is bound to the run, its repository, | |
| 656 | + | and what that kind of run needs. It expires no later than the run's | |
| 657 | + | timeout. | |
| 658 | + | - **What you may do now.** It works only in the repository's workspace, | |
| 659 | + | with your [role](/guides/access-and-roles/) on the repository as it is now, and never | |
| 660 | + | more than Write: an owner's agent has Write, not Admin. If you leave the | |
| 661 | + | workspace or lose your role, every agent working on your behalf there | |
| 662 | + | loses it too. It can never change who has access. | |
| 663 | + | ||
| 664 | + | A sandbox holds two credentials. One is for g1t's runner, which clones, | |
| 665 | + | pushes the result and records the session; downstream it acts as you, so | |
| 666 | + | what it pushes is yours, within the run's scope. The other is for the | |
| 667 | + | agent's own tools over MCP, and acts as the agent; it cannot be used with | |
| 668 | + | git at all. | |
| 669 | + | ||
| 670 | + | | Kind of run | Git | API and MCP tools | | |
| 671 | + | | --- | --- | --- | | |
| 672 | + | | Implement | Reads the repository; pushes to its pull request's fork only | Records the session and marks its own pull request ready; tools to read issues, pull requests, the merge queue, workflow runs and memory, to [search all of g1t](/guides/search/) and the workspace's context hub, open issues, comment, remember, and message other agents | | |
| 673 | + | | Revise, answer | Reads the repository; pushes to the pull request's fork, or to its branch only when the change is a branch of the repository | Records the session of its own pull request; the same tools as implement | | |
| 674 | + | | Catch up | Reads the repository; pushes to the pull request's fork or branch only | Records the session of its own pull request | | |
| 675 | + | | Review | Reads the change and the repository; pushes nothing | Reports its review through its own run | | |
| 676 | + | | Plan | Reads the repository; pushes nothing | Reports its plan through its own run, for a person to apply; it can create issues in its repository only | | |
| 677 | + | | Merge check | Reads the change; pushes nothing | None | | |
| 678 | + | | Merge queue | Reads each queued change; pushes the queue's own branch only | None | | |
| 679 | + | | Deploy | Reads the commit it builds; pushes nothing | None | | |
| 680 | + | ||
| 681 | + | Nothing an agent's credential holds can reach another repository, or a | |
| 682 | + | workspace's settings, members, access tokens, billing, integrations, | |
| 683 | + | webhooks, secrets and variables, or workflows' controls. It cannot merge a | |
| 684 | + | pull request or put more agents to work. It cannot change its repository's | |
| 685 | + | details or default branch, rename it or its branches, make it public or | |
| 686 | + | private, archive, transfer, delete, restore or purge it; see | |
| 687 | + | [managing a repository](/guides/managing-repositories/). A call that would is refused, and | |
| 688 | + | the refusal is recorded with the rule that refused it: | |
| 689 | + | ||
| 690 | + | | Rule | Refused because | | |
| 691 | + | | --- | --- | | |
| 692 | + | | `never` | No agent's credential may ever do this. | | |
| 693 | + | | `scope:operation` | The run's kind does not include this operation. | | |
| 694 | + | | `scope:repository` | It names a repository other than the run's. | | |
| 695 | + | | `scope:pull` | The runner tried to change a pull request other than its own. | | |
| 696 | + | | `on-behalf-of:membership` | The person the agent works for is no longer a member of the workspace, and has no role on its repositories. | | |
| 697 | + | | `git:read`, `git:push`, `git:ref` | The run has no grant to clone that repository, push to it, or move that branch or tag. | | |
| 698 | + | | `git:not-a-run` | An agent's tools credential was used with git. | | |
| 699 | + | ||
| 700 | + | Personal and workspace access tokens are unchanged by any of this. | |
| 701 | + | ||
| 702 | + | ## What a sandbox can reach | |
| 703 | + | ||
| 704 | + | A sandbox holds one fork and its run's credentials, which expire when the | |
| 705 | + | run's time is up and are revoked as soon as it stops. Those credentials, | |
| 706 | + | and the model key the agent runs on, are removed from anything recorded in | |
| 707 | + | the session. | |
| 708 | + | ||
| 709 | + | ## Guardrails | |
| 710 | + | ||
| 711 | + | A workspace decides what its agents may do in their sandboxes, and each | |
| 712 | + | project can override it: which hosts a sandbox can reach (g1t, the package | |
| 713 | + | registries the project needs, and domains you list; enforced outside the | |
| 714 | + | sandbox), which commands the harness refuses (force-pushing, rewriting the | |
| 715 | + | default branch, reading outside the project, printing the environment, | |
| 716 | + | sudo, and your own patterns), and how much one run may cost and how long it | |
| 717 | + | may take. A run that is refused something shows it as a step; one that | |
| 718 | + | reaches a cap is stopped and its pull request waits for you. A run on a | |
| 719 | + | fork's head loads none of the fork's `CLAUDE.md`, `.claude` settings, | |
| 720 | + | hooks, MCP servers or commands. See [guardrails](/guides/guardrails/) for | |
| 721 | + | every rule and exactly how each is enforced. |
| 35 | 35 | Usernames and workspaces share one set of names, so a name means the same | |
| 36 | 36 | thing wherever it appears. Your username is reserved for you: only you can | |
| 37 | 37 | create a workspace with that name, and nobody can register a username that | |
| 38 | − | is already a workspace. | |
| 38 | + | is already a workspace. The names `g1t` and `g1t-agent` belong to | |
| 39 | + | [g1t's agent](/guides/working-with-g1t/), and nobody can register them. | |
| 39 | 40 | ||
| 40 | 41 | ## Display name, slug and icon | |
| 41 | 42 | ||
| 211 | 212 | ||
| 212 | 213 | | Role | Can | | |
| 213 | 214 | | --- | --- | | |
| 214 | − | | Member | Create repositories, see the workspace's usage and billing, and get the workspace's [base permission](/guides/access-and-roles/#the-base-permission) on every repository in it: Write unless an owner changes it, which is enough to push, manage issues, merge pull requests, plan work and put g1t agents to work. | | |
| 215 | + | | Member | Create repositories, see the workspace's usage and billing, and get the workspace's [base permission](/guides/access-and-roles/#the-base-permission) on every repository in it: Write unless an owner changes it, which is enough to push, manage issues, merge pull requests, plan work and put g1t to work. | | |
| 215 | 216 | | Owner | Everything a member can, and manage members, the base permission, the workspace's access tokens, its details, and billing: the plan, card checks, prepayment and limits. Admin on every repository, and the only ones who can transfer and delete them; see [access and roles](/guides/access-and-roles/). | | |
| 216 | 217 | ||
| 217 | 218 | Whoever creates a workspace is its owner. An owner adds people on the | |
| 254 | 255 | across them (**Agent fleet**, **Context**, **Memory**, **Security**, | |
| 255 | 256 | **Guardrails**, [**Secrets and variables**](/guides/secrets-and-variables/), | |
| 256 | 257 | [**Integrations**](/guides/integrations/) and | |
| 257 | − | [**Webhooks**](/guides/webhooks/)); then **Usage**, what g1t agents have | |
| 258 | + | [**Webhooks**](/guides/webhooks/)); then **Usage**, what g1t's runs have | |
| 258 | 259 | cost (see [usage and billing](/guides/usage-and-billing/)), **Support** and | |
| 259 | 260 | **Settings**. An item with an arrow opens a list of its own in the sidebar: | |
| 260 | 261 | **Settings** slides over to the workspace's settings, and the row at the | |
| 372 | 373 | ||
| 373 | 374 | - **Overview:** pull requests merged, open pull requests and issues | |
| 374 | 375 | opened, and your most recent activity. | |
| 375 | − | - **Pull requests** and **Issues:** everything you opened, newest first, | |
| 376 | + | - **Pull requests** and **Issues:** everything you opened, and what g1t | |
| 377 | + | opened for you, newest first, | |
| 376 | 378 | with filters beside the list for state (open, closed, merged), type, | |
| 377 | 379 | repository and sort order. Add `?tab=pulls&state=merged` and the like to | |
| 378 | 380 | link to a filtered list. |
| 85 | 85 | ||
| 86 | 86 | <Tabs> | |
| 87 | 87 | <TabItem label="g1t's agent"> | |
| 88 | − | On the issue, choose **Assign to g1t agent**. It opens a pull | |
| 88 | + | On the issue, choose **Assign to g1t**. It opens a pull | |
| 89 | 89 | request, makes the change in a sandbox, and sees it through: checks and | |
| 90 | 90 | workflows, a review by a second agent, revisions until both pass. | |
| 91 | 91 | </TabItem> | |
| 113 | 113 | ## Next | |
| 114 | 114 | ||
| 115 | 115 | - [How g1t works](/concepts/overview/): issues, pull requests, checks, review and merging. | |
| 116 | − | - [Hand off an outcome](/guides/outcomes/): an agent plans the issues and g1t agents land them in order. | |
| 116 | + | - [Hand off an outcome](/guides/outcomes/): an agent plans the issues and g1t lands them in order. | |
| 117 | 117 | - [GitHub Actions](/guides/actions/): your workflows, unchanged. | |
| 118 | 118 | - [MCP tools](/reference/mcp/) and the [API](/reference/api/): everything above, from code. |
| 52 | 52 | says whether it needs a token. | |
| 53 | 53 | ||
| 54 | 54 | A [workspace's own token](/guides/workspaces/#workspace-access-tokens) acts | |
| 55 | − | as the workspace. The token a g1t agent works with can use only the | |
| 55 | + | as the workspace. The token g1t works with can use only the | |
| 56 | 56 | operations its task needs, in its own repository. | |
| 57 | 57 | ||
| 58 | 58 | ### Scopes | |
| 198 | 198 | | [Invites](/reference/api/invites/list-invites/) | Your invites while g1t is invite-only, and inviting people into a workspace by email. | | |
| 199 | 199 | | [Repositories](/reference/api/repositories/list-repos/) | A repository, how it handles pull requests, and its timeline. | | |
| 200 | 200 | | [Access](/reference/api/access/list-collaborators/) | Who has which role on a repository, invitations, outside collaborators, and a workspace's base permission. | | |
| 201 | − | | [Issues](/reference/api/issues/list-issues/) | What should change, with labels and comments, and assigning it to the g1t agent. | | |
| 201 | + | | [Issues](/reference/api/issues/list-issues/) | What should change, with labels and comments, and assigning it to g1t. | | |
| 202 | 202 | | [Plans](/reference/api/plans/plan-work/) | An [outcome](/guides/outcomes/) turned into issues. | | |
| 203 | 203 | | [Pull requests](/reference/api/pull-requests/list-pull-requests/) | Proposed changes: reviews, merging, the merge queue, and messages to the agent at work. | | |
| 204 | 204 | | [Sessions](/reference/api/sessions/read-session/) | The record of how a pull request was made. | |
| 236 | 236 | | [`create`](/reference/api/repositories/create-repo/) | Create a repository in one of your workspaces, empty or as a copy of a public git repository (`import_url`). `workspace` may be left out if you belong to exactly one. | `name` | `repo:write` | | |
| 237 | 237 | | [`update`](/reference/api/repositories/update-repo/) | Change its `description`, `website`, `topics` and `default_branch`, whether its default branch is `protected`, and whether it is `private`. Maintain role; `private` and `default_branch` need Admin. | `repo` | `repo:write` | | |
| 238 | 238 | | [`get_settings`](/reference/api/repositories/get-repo-settings/) | How it handles pull requests: the default branch's required checks, approvals, bypassing checks, being up to date, the merge queue, and how g1t's agents are reviewed, revised and merged. | `repo` | `repo:read` | | |
| 239 | − | | [`update_settings`](/reference/api/repositories/update-repo-settings/) | Change those settings, including `hold_low_confidence`, which holds a g1t agent's [low-confidence](/guides/g1t-agents/#how-sure-the-agent-is) change for a person. Only the fields given change; `required_checks` replaces the whole list. Maintain role. | `repo` | `repo:write` | | |
| 239 | + | | [`update_settings`](/reference/api/repositories/update-repo-settings/) | Change those settings, including `hold_low_confidence`, which holds g1t's [low-confidence](/guides/working-with-g1t/#how-sure-the-agent-is) change for a person. Only the fields given change; `required_checks` replaces the whole list. Maintain role. | `repo` | `repo:write` | | |
| 240 | 240 | | [`check_names`](/reference/api/repositories/list-check-names/) | The check names reported on its commits in the last 30 days, most recent first, each with `name`, `events` and `last_seen`: the names `required_checks` takes. | `repo` | `repo:read` | | |
| 241 | 241 | | [`list_labels`](/reference/api/issues/list-labels/) | The labels available on its issues. | `repo` | `repo:read` | | |
| 242 | 242 | | [`list_events`](/reference/api/repositories/list-events/) | Its timeline, newest first. `before` pages back. | `repo` | `repo:read` | | |
| 259 | 259 | `allow_ignoring_checks`, `require_up_to_date`, `agent_review`, | |
| 260 | 260 | `max_revisions`, `auto_merge`, `merge_queue` and `hold_low_confidence`. See | |
| 261 | 261 | [required status checks](/guides/pull-requests/#required-status-checks) and | |
| 262 | − | [what a repository can ask for](/guides/g1t-agents/#what-a-repository-can-ask-for). | |
| 262 | + | [what a repository can ask for](/guides/working-with-g1t/#what-a-repository-can-ask-for). | |
| 263 | 263 | `update` with `private` or `default_branch` also needs `repo:admin`. | |
| 264 | 264 | ||
| 265 | 265 | See [managing a repository](/guides/managing-repositories/) for what each | |
| 281 | 281 | | [`close`](/reference/api/issues/close-issue/) | Close it as `completed` or `not_planned`. | `repo`, `number` | `issues:write` | | |
| 282 | 282 | | [`reopen`](/reference/api/issues/reopen-issue/) | Reopen a closed issue. | `repo`, `number` | `issues:write` | | |
| 283 | 283 | | [`comment`](/reference/api/issues/add-comment/) | Comment on an issue or a pull request; with `path` and `line`, on one line of a pull request's change. | `repo`, `number`, `body` | `issues:write` | | |
| 284 | − | | [`import`](/reference/api/integrations/import-issue/) | Open an issue from a ticket, linked to it. `assign` puts a g1t agent on it. | `repo`, `reference` | `issues:write` | | |
| 284 | + | | [`import`](/reference/api/integrations/import-issue/) | Open an issue from a ticket, linked to it. `assign` assigns it to g1t. | `repo`, `reference` | `issues:write` | | |
| 285 | 285 | ||
| 286 | 286 | `import` with `assign` also needs `agents:run`, since it puts an agent to | |
| 287 | 287 | work. | |
| 301 | 301 | | [`record_session`](/reference/api/sessions/record-session/) | Append entries to a pull request's session. Each has `kind` and `text`, and `tool` for tool entries. | `repo`, `number`, `entries` | `pull_requests:write` | | |
| 302 | 302 | | [`read_session`](/reference/api/sessions/read-session/) | The recorded session, oldest first. `after` skips to entries after a sequence number. | `repo`, `number` | `pull_requests:read` | | |
| 303 | 303 | | [`ready`](/reference/api/pull-requests/mark-pull-request-ready/) | Mark a draft ready for review. The summary becomes its description. | `repo`, `number`, `summary` | `pull_requests:write` | | |
| 304 | − | | [`review`](/reference/api/pull-requests/review-pull-request/) | `approve`, or `request_changes` with a `body`. Not on your own pull request. | `repo`, `number`, `verdict` | `pull_requests:write` | | |
| 304 | + | | [`review`](/reference/api/pull-requests/review-pull-request/) | `approve`, or `request_changes` with a `body`. Not on your own pull request, nor one g1t made for you. | `repo`, `number`, `verdict` | `pull_requests:write` | | |
| 305 | 305 | | [`close`](/reference/api/pull-requests/close-pull-request/) | Close it without merging. | `repo`, `number` | `pull_requests:write` | | |
| 306 | 306 | | [`merge`](/reference/api/pull-requests/merge-pull-request/) | Land it on `main` and resolve its issue, or add it to the [merge queue](/guides/merge-queue/), once every [required check](/guides/pull-requests/#required-status-checks) has passed on its head. `ignore_checks` bypasses them where the repository allows it. Write role. | `repo`, `number` | `pull_requests:write` | | |
| 307 | 307 | | [`merge_queue`](/reference/api/pull-requests/get-merge-queue/) | The pull requests waiting to land, in order, each with the state it is tested in and how that went; then those that recently landed or left. | `repo` | `pull_requests:read` | | |
| 313 | 313 | ||
| 314 | 314 | ## `agent` | |
| 315 | 315 | ||
| 316 | − | Put [g1t agents](/guides/g1t-agents/) to work and talk to them. One agent | |
| 316 | + | Put [g1t](/guides/working-with-g1t/) to work and talk to it. One run | |
| 317 | 317 | works on each issue; to do more at once, use more issues. Starting an agent | |
| 318 | 318 | uses the workspace's money. `delegate` also needs `issues:write`, since it | |
| 319 | 319 | opens the issue. | |
| 320 | 320 | ||
| 321 | 321 | | Action | What it does | Required | Scope | | |
| 322 | 322 | | --- | --- | --- | --- | | |
| 323 | − | | [`delegate`](/reference/api/issues/delegate/) | Put an agent on something in one step: open an issue, with `body`, and assign it to the g1t agent at once. Write role; nothing is opened without it. The issue opens even when the agent cannot start: `agent.status` is `started`, `queued` or `not_started`, with `agent.code`, `agent.message` and `agent.fix_url` saying why and where to fix it. `checks` is deprecated, as for `issue` `create`. See [put an agent on it](/guides/g1t-agents/#put-an-agent-on-it-in-one-step). | `repo`, `title` | `agents:run` | | |
| 324 | − | | [`assign`](/reference/api/issues/assign-issue/) | Assign an existing issue to the [g1t agent](/guides/g1t-agents/), which opens a pull request and sees it through. Preview. | `repo`, `number` | `agents:run` | | |
| 325 | − | | [`message`](/reference/api/pull-requests/message-agent/) | Send the agent working on a pull request a message, received at its next step. A g1t agent sends a `question` or a `handoff`, with its own pull request as `from_number`. | `repo`, `number`, `body` | `agents:run` | | |
| 323 | + | | [`delegate`](/reference/api/issues/delegate/) | Put an agent on something in one step: open an issue, with `body`, and assign it to g1t at once. Write role; nothing is opened without it. The issue opens even when the agent cannot start: `agent.status` is `started`, `queued` or `not_started`, with `agent.code`, `agent.message` and `agent.fix_url` saying why and where to fix it. `checks` is deprecated, as for `issue` `create`. See [put an agent on it](/guides/working-with-g1t/#put-an-agent-on-it-in-one-step). | `repo`, `title` | `agents:run` | | |
| 324 | + | | [`assign`](/reference/api/issues/assign-issue/) | Assign an existing issue to [g1t](/guides/working-with-g1t/), which opens a pull request and sees it through. Preview. | `repo`, `number` | `agents:run` | | |
| 325 | + | | [`message`](/reference/api/pull-requests/message-agent/) | Send the agent working on a pull request a message, received at its next step. g1t sends a `question` or a `handoff`, with its own pull request as `from_number`. | `repo`, `number`, `body` | `agents:run` | | |
| 326 | 326 | | [`answer`](/reference/api/pull-requests/answer-message/) | Answer a question or a handoff by the message's id; `decline` a handoff that is not yours. The answer reaches the asking agent at its next step. | `repo`, `id`, `body` | `agents:run` | | |
| 327 | − | | [`take_messages`](/reference/api/pull-requests/take-messages/) | For a g1t agent at work: the messages it has not seen yet, each returned once. | `repo`, `number` | `agents:run` | | |
| 327 | + | | [`take_messages`](/reference/api/pull-requests/take-messages/) | For g1t at work: the messages it has not seen yet, each returned once. | `repo`, `number` | `agents:run` | | |
| 328 | 328 | ||
| 329 | 329 | See [talk to agents](/guides/talking-to-agents/). | |
| 330 | 330 | ||
| 338 | 338 | | --- | --- | --- | --- | | |
| 339 | 339 | | [`create`](/reference/api/plans/plan-work/) | Have an agent turn an outcome into proposed issues, each with what done means (`done`) and its dependencies. Returns the plan's id at once. Write role. | `repo`, `brief` | `agents:run` | | |
| 340 | 340 | | [`get`](/reference/api/plans/get-plan/) | The plan: its status (`planning`, `ready`, `failed` or `applied`), the issues it proposes, and once applied, where each stands. | `repo`, `plan` | `issues:read` | | |
| 341 | − | | [`apply`](/reference/api/plans/apply-plan/) | Open its issues. `assign` puts g1t agents on them in dependency order; `keep` opens only some, by position from 1. | `repo`, `plan` | `issues:write` | | |
| 341 | + | | [`apply`](/reference/api/plans/apply-plan/) | Open its issues. `assign` assigns them to g1t in dependency order; `keep` opens only some, by position from 1. | `repo`, `plan` | `issues:write` | | |
| 342 | 342 | ||
| 343 | 343 | ## `memory` | |
| 344 | 344 | ||
| 473 | 473 | | [`decline_repository_invitation`](/reference/api/access/decline-repo-invitation/) | Decline one. | `id` | `account:write` | | |
| 474 | 474 | ||
| 475 | 475 | ||
| 476 | − | ## What a g1t agent can use | |
| 476 | + | ## What g1t can use | |
| 477 | 477 | ||
| 478 | − | A g1t agent works with a [run credential](/guides/g1t-agents/#credentials): | |
| 479 | − | a token bound to its run and its own repository, acting as `g1t-agent` on | |
| 478 | + | g1t works with a [run credential](/guides/working-with-g1t/#credentials): | |
| 479 | + | a token bound to its run and its own repository, acting as `g1t` on | |
| 480 | 480 | behalf of the person who started the work, and only while that person is | |
| 481 | 481 | still a member of the workspace or has a role on one of its repositories. It | |
| 482 | 482 | has that person's role on its repository, but never more than Write. Which |
Binary or large file; its contents are not shown.
| 16 | 16 | ||
| 17 | 17 | import { Avatar, TimeAgo } from "./ui"; | |
| 18 | 18 | ||
| 19 | − | /** The accounts that are g1t's own agents and machinery. */ | |
| 20 | − | const AGENTS = new Set(["g1t-agent", "g1t"]); | |
| 19 | + | /** Whether an actor is g1t itself: its agent or its machinery. */ | |
| 20 | + | const isG1t = (actor: string | null): actor is string => actor === "g1t"; | |
| 21 | 21 | ||
| 22 | 22 | type Line = { icon: ReactNode; tone: string; actor: string | null; text: ReactNode; coordination?: boolean }; | |
| 23 | 23 | ||
| 33 | 33 | case "issue.opened": | |
| 34 | 34 | return { | |
| 35 | 35 | icon: <CircleDot size={14} />, | |
| 36 | − | tone: actor && AGENTS.has(actor) ? "text-merged" : "text-muted", | |
| 36 | + | tone: isG1t(actor) ? "text-merged" : "text-muted", | |
| 37 | 37 | actor, | |
| 38 | − | text: actor && AGENTS.has(actor) ? <>filed {ref(event.data.number)} for something it found</> : <>opened {ref(event.data.number)}</>, | |
| 39 | − | coordination: actor != null && AGENTS.has(actor), | |
| 38 | + | text: isG1t(actor) ? <>filed {ref(event.data.number)} for something it found</> : <>opened {ref(event.data.number)}</>, | |
| 39 | + | coordination: isG1t(actor), | |
| 40 | 40 | }; | |
| 41 | 41 | case "pull.opened": | |
| 42 | 42 | return { | |
| 66 | 66 | return { | |
| 67 | 67 | icon: event.data.verdict === "approve" ? <CircleCheck size={14} /> : <CircleSlash size={14} />, | |
| 68 | 68 | tone: event.data.verdict === "approve" ? "text-accent" : "text-warn", | |
| 69 | − | actor: "g1t-agent", | |
| 69 | + | actor: "g1t", | |
| 70 | 70 | text: ( | |
| 71 | 71 | <> | |
| 72 | 72 | {event.data.verdict === "approve" ? "approved" : "asked for changes on"} {ref(event.data.number)} | |
| 88 | 88 | } | |
| 89 | 89 | return { | |
| 90 | 90 | icon: <MessageSquare size={14} />, | |
| 91 | − | tone: actor && AGENTS.has(actor) ? "text-merged" : "text-muted", | |
| 91 | + | tone: isG1t(actor) ? "text-merged" : "text-muted", | |
| 92 | 92 | actor, | |
| 93 | 93 | text: <>commented on {ref(event.data.number)}</>, | |
| 94 | − | coordination: actor != null && AGENTS.has(actor), | |
| 94 | + | coordination: isG1t(actor), | |
| 95 | 95 | }; | |
| 96 | 96 | case "pull.merged": | |
| 97 | 97 | return { icon: <GitMerge size={14} />, tone: "text-accent", actor, text: <>landed {ref(event.data.number)} on main</> }; | |
| 161 | 161 | return ( | |
| 162 | 162 | <li key={exchange.id} className="rounded-2xl bg-merged/[0.05] p-4 ring-1 ring-merged/25"> | |
| 163 | 163 | <p className="flex flex-wrap items-center gap-x-2 gap-y-1 text-xs text-muted"> | |
| 164 | − | <Avatar name="g1t-agent" size={14} /> | |
| 164 | + | <Avatar name="g1t" size={14} /> | |
| 165 | 165 | <span> | |
| 166 | 166 | The agent on {exchange.fromNumber != null ? pull(exchange.fromNumber) : exchange.author}{" "} | |
| 167 | 167 | {exchange.kind === "handoff" ? "handed work to" : "asked"} the agent on {pull(exchange.toNumber)} |
| 14 | 14 | ||
| 15 | 15 | /** | |
| 16 | 16 | * "Put an agent on it": a compact composer that opens an issue in one of | |
| 17 | − | * the viewer's projects and puts g1t-agent on it, in one step. Posts to | |
| 17 | + | * the viewer's projects and puts g1t on it, in one step. Posts to | |
| 18 | 18 | * Mission control's action, which lands on the issue with the agent | |
| 19 | 19 | * running. A native `<details>`, so it opens and posts without script. | |
| 20 | 20 | */ | |
| 79 | 79 | Put an agent on it | |
| 80 | 80 | </p> | |
| 81 | 81 | <p className="mt-1 text-xs leading-5 text-muted"> | |
| 82 | − | Opens an issue and assigns g1t-agent at once. It makes the change in a sandbox and sees it through checks | |
| 82 | + | Opens an issue and assigns g1t at once. It makes the change in a sandbox and sees it through checks | |
| 83 | 83 | and review. | |
| 84 | 84 | </p> | |
| 85 | 85 | ||
| 129 | 129 | <div className="mt-3 rounded-lg border border-warn/30 bg-warn/[0.06] p-3 text-sm"> | |
| 130 | 130 | <p className="text-fg-soft"> | |
| 131 | 131 | Opened <Link to={said.notStarted.to} className="font-medium text-fg hover:underline">#{said.notStarted.number}</Link>, but | |
| 132 | − | g1t-agent did not start. {said.notStarted.message} | |
| 132 | + | g1t did not start. {said.notStarted.message} | |
| 133 | 133 | </p> | |
| 134 | 134 | <p className="mt-2 flex flex-wrap gap-2"> | |
| 135 | 135 | {said.notStarted.fix && ( | |
| 150 | 150 | ||
| 151 | 151 | <div className="mt-4 flex items-center justify-between gap-3"> | |
| 152 | 152 | <span className="flex min-w-0 items-center gap-1.5 text-xs text-faint"> | |
| 153 | − | <Avatar name="g1t-agent" size={16} /> | |
| 154 | − | <span className="truncate">g1t-agent · no model to choose</span> | |
| 153 | + | <Avatar name="g1t" size={16} /> | |
| 154 | + | <span className="truncate">g1t · no model to choose</span> | |
| 155 | 155 | </span> | |
| 156 | 156 | <button | |
| 157 | 157 | type="submit" |
| 13 | 13 | }; | |
| 14 | 14 | ||
| 15 | 15 | /** | |
| 16 | − | * The files every g1t agent run in a project reads as the repository's | |
| 16 | + | * The files every run of g1t in a project reads as the repository's | |
| 17 | 17 | * instructions, as they are on its default branch: which, what they say, | |
| 18 | 18 | * when they last changed, and where to change them. | |
| 19 | 19 | */ |
| 252 | 252 | </Link> | |
| 253 | 253 | <p className="mt-1.5 flex flex-wrap items-center gap-x-2 gap-y-1 text-xs text-muted"> | |
| 254 | 254 | {current.pull.runtime === "hosted" ? ( | |
| 255 | − | <span className="flex items-center gap-1 text-merged"> | |
| 256 | − | <Sparkles size={11} /> | |
| 257 | − | written by {current.pull.agent} | |
| 258 | − | </span> | |
| 255 | + | <> | |
| 256 | + | <span className="flex items-center gap-1 text-merged"> | |
| 257 | + | <Sparkles size={11} /> | |
| 258 | + | written by {current.pull.agent} | |
| 259 | + | </span> | |
| 260 | + | {current.pull.requestedBy && <span>for {current.pull.requestedBy}</span>} | |
| 261 | + | </> | |
| 259 | 262 | ) : ( | |
| 260 | 263 | <span>by {current.pull.author} with {current.pull.agent}</span> | |
| 261 | 264 | )} |
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
Binary or large file; its contents are not shown.
This change is too large to show in full.