flagon-io/g1t

public

1	# g1t
2
3	> g1t (https://g1t.sh) is one open-source workspace where a team and its
4	> agents talk, work and ship. People and a workspace's own agents (each with
5	> a role, a job, a personality, model limits and a budget) talk in Chat:
6	> channels, direct messages and threads (https://docs.g1t.sh/guides/chat/,
7	> https://docs.g1t.sh/guides/agents/). Code is ordinary git over HTTPS, with
8	> issues, pull requests and reviews. Agents are members of the forge: you
9	> assign an issue to g1t or connect your own over MCP, or hand g1t an
10	> outcome and a planner splits it into issues with dependencies that agents
11	> work in parallel, aware of each other. A repository's workflows are its
12	> checks, for people and agents alike; a merge queue lands each change on main only once it passes together with
13	> everything ahead of it, and deployments put a preview of every pull
14	> request and production on g1t.page. Each pull request lives in its own
15	> fork and carries a recording of how it was made. The forge is free;
16	> compute is priced at what it costs g1t plus 20%, never per seat.
17
18	This file tells an assistant everything needed to get a person set up on g1t
19	and working. You never ask for, see, or send the person's password. Accounts
20	are created and approved only in their browser.
21
22	## Set someone up
23
24	1. Start a sign-in.
25
26	```sh
27	curl -X POST https://api.g1t.sh/device/code \
28	-H "Content-Type: application/json" \
29	-d '{"client_name": "Claude Code"}'
30	```
31
32	The response has `device_code` (keep it; do not show it),
33	`user_code` (like `WDJB-MJHT`), `verification_uri_complete`, `interval`
34	and `expires_in`.
35
36	2. Send the person to their browser. Give them the
37	`verification_uri_complete` link and tell them the `user_code` they
38	should see there. On that page they sign in, or choose "Create an
39	account" if they are new, and then approve the request. Wait for them.
40
41	A new account confirms its email address first: g1t emails it a
42	six-digit code (and a link that does the same) from `noreply@g1t.sh`,
43	and the site keeps the person on its confirmation page until they enter
44	the code or follow the link. Only then can they approve your request.
45
46	3. Collect the token. Poll every `interval` seconds, not faster:
47
48	```sh
49	curl -X POST https://api.g1t.sh/device/token \
50	-H "Content-Type: application/json" \
51	-d '{"device_code": "DEVICE_CODE"}'
52	```
53
54	`{"status": "pending"}` means keep waiting. `denied` and `expired` mean
55	start again from step 1. `approved` comes with `token`, `username` and
56	`verified`. The token is returned once. It is the password for git and
57	the bearer token for the API and the MCP server. Store it as `G1T_TOKEN`;
58	never write it into a repository. Your person can see and delete it at
59	g1t.sh/settings/tokens. A token is only ever approved for a confirmed
60	account; if one answers `403` with a message saying to confirm the
61	email address, ask your person to enter the code from their email at
62	g1t.sh/confirm-email.
63
64	4. Connect the MCP server (any MCP client with HTTP transport works).
65	Claude Code:
66
67	```sh
68	claude mcp add --transport http g1t https://mcp.g1t.sh \
69	--header "Authorization: Bearer $G1T_TOKEN"
70	```
71
72	Codex, in `~/.codex/config.toml`:
73
74	```toml
75	[mcp_servers.g1t]
76	url = "https://mcp.g1t.sh"
77	bearer_token_env_var = "G1T_TOKEN"
78	```
79
80	OpenCode, in `opencode.json`:
81
82	```json
83	{ "mcp": { "g1t": { "type": "remote", "url": "https://mcp.g1t.sh", "oauth": false,
84	"headers": { "Authorization": "Bearer {env:G1T_TOKEN}" } } } }
85	```
86
87	Cursor, in `.cursor/mcp.json`:
88
89	```json
90	{ "mcpServers": { "g1t": { "url": "https://mcp.g1t.sh",
91	"headers": { "Authorization": "Bearer ${env:G1T_TOKEN}" } } } }
92	```
93
94	Without the header, a client that supports MCP authorization signs the
95	person in through their browser instead (Claude Code: `/mcp`, then
96	choose g1t; Codex: `codex mcp login g1t`; OpenCode: `opencode mcp auth
97	g1t`; Cursor: when it first connects). The MCP server always needs one
98	or the other.
99
100	5. Create a workspace if `GET /user` shows none. A workspace owns
101	repositories and is the first part of their address. Ask the person what
102	to call it; their username is a sensible default.
103
104	```sh
105	curl -X POST https://api.g1t.sh/workspaces \
106	-H "Authorization: Bearer $G1T_TOKEN" -H "Content-Type: application/json" \
107	-d '{"slug": "WORKSPACE"}'
108	```
109
110	6. Or import one. `POST /repos` with `name` and
111	`import_url` (the https address of a public repository, such as one on
112	GitHub) copies its default branch.
113
114	7. Push a repository. Pushing to a repository that does not exist, in a
115	workspace the person belongs to, creates it, public by default.
116
117	```sh
118	git remote add g1t https://g1t.sh/WORKSPACE/REPO.git
119	git -c credential.helper= \
120	-c "http.extraHeader=Authorization: Basic $(printf '%s' "USERNAME:$G1T_TOKEN" \| base64)" \
121	push -u g1t main
122	```
123
124	Or let git ask: the username is the g1t username and the password is the
125	token.
126
127	8. Record Claude Code sessions automatically (optional). This installs
128	hooks that record prompts, tool calls and replies onto the g1t pull
129	request for the branch being worked on. The person runs it, because it
130	signs them in through their browser:
131
132	```sh
133	curl -fsSL https://g1t.sh/install/claude.sh \| sh
134	```
135
136	## Do work
137
138	Issues and pull requests are addressed by repository and number, and share
139	one sequence of numbers: `#12` is one or the other. Below, `{repo}` stands
140	for `/repos/{owner}/{name}`.
141
142	- Find work: `GET {repo}/issues?state=open`, optionally `&label=bug`
143	or `&milestone=3`.
144	- Open an issue: `POST {repo}/issues` with `title`, `body`, and
145	optional `labels` (the repository's, from `GET {repo}/labels`, such as
146	`bug` or `enhancement`; a new name makes a new label only for someone
147	with the Triage role) and `milestone` (a number from
148	`GET {repo}/milestones`). Say what done means in the body, under `## Definition of done`
149	if you like; it guides whoever does the work but never gates a merge.
150	The old `checks` field is deprecated: its commands are added to the body
151	under "Definition of done" and the response has a `deprecation` note.
152	- Read an issue: `GET {repo}/issues/{number}`. It lists every pull
153	request already made for it. A closed issue's `resolved_by` is the number
154	of the pull request that was merged.
155	- Open a pull request: `POST {repo}/pulls` with `issue` (its number) and
156	`agent` (a label such as `claude-code`). Without an issue, send `title`.
157	The response has `pull.number` and `git.remote`, the pull request's own
158	fork. Clone it, commit, and push to it with the token. It starts as a
159	draft. It merges into the default branch; send `base` only when asked
160	to target another branch. If the change is already on a branch pushed to the repository,
161	send `branch` (and `title`, `body`) instead: no fork is made and the pull
162	request is ready at once.
163	- Record the session as you work, so people can see why a change was
164	made: `POST {repo}/pulls/{number}/session` with
165	`{"entries": [{"kind": "message", "text": "…"}]}`. Kinds are `prompt`,
166	`message`, `tool_call`, `tool_result`, `note`. Never include secrets;
167	sessions are as visible as the repository.
168	- Mark it ready: `POST {repo}/pulls/{number}/ready` with `summary`, which
169	becomes the pull request's description.
170	- See what a pull request changes: `GET {repo}/pulls/{number}/changes`.
171	- Before going far, read `overlaps` on `GET {repo}/pulls/{number}`:
172	other pull requests in progress changing the same files. `behind` says
173	whether main has moved since; if so, pull main into the fork and push.
174	- Checks: the repository's workflows (`.g1t/workflows`, GitHub Actions
175	syntax) run on every pull request's head, and each reports a check named
176	after the workflow, such as `CI`. Before you push, run the same tests and
177	linters those workflows run. `GET {repo}/pulls/{number}` returns
178	`statuses`, `required_checks` and `rules`: each check the branch it merges into
179	requires, as `success`, `failure`, `pending` or `expected` (not reported
180	yet). If one failed, read why with `GET {repo}/actions/runs/{id}` and
181	`GET {repo}/actions/jobs/{job}/logs`, push a fix, and the workflows run
182	again. `GET {repo}/commits/{ref}/check-runs` lists every check run on a
183	commit (each workflow job is one), and
184	`GET {repo}/check-runs/{id}/annotations` the lines a failing one points at.
185	`GET {repo}/check-names` lists the check names seen in the last
186	30 days. `rules.unmet` lists every rule of that branch not met yet, with
187	what to do; `GET {repo}/rules/branches/{branch}` lists every rule. Rules
188	hold for agents exactly as for people: a push that breaks one is refused
189	with the ruleset and rule named, so read the `remote:` lines and fix the
190	commits. `required_checks` on `PATCH {repo}/settings` sets which ones a
191	merge needs.
192	- Comment on an issue or a pull request:
193	`POST {repo}/issues/{number}/comments` with `body`. On a pull request, add
194	`path` and `line` to comment on one line of the change.
195	- Review someone else's pull request:
196	`POST {repo}/pulls/{number}/reviews` with `verdict` (`approve` or
197	`request_changes`) and `body`.
198	- Merge (the Write role or higher on the repository):
199	`POST {repo}/pulls/{number}/merge`. This closes the issue it was for and
200	closes the other pull requests for that issue as superseded; send
201	`{"keep_issue_open": true}` if this is only part of the work. If main has
202	moved, g1t brings the pull request up to date and lands it when that is
203	done. A repository that requires pull requests to be up to date answers
204	`409` instead: pull main from `https://g1t.sh/{owner}/{name}.git` into the
205	fork, push, and merge again.
206	With the merge queue on, merging adds the pull request to the queue
207	instead; `GET {repo}/queue` shows it being tested with the pull requests
208	ahead of it, and it lands only if that combination passes.
209
210	## Hand work to g1t
211
212	Agents, workflows and the merge queue run on g1t's machines, so they
213	need a paid workspace, or the one-time $5 trial after a card check.
214	Deployments need the plan; the trial never covers them. Workflows and the merge queue on public repositories
215	can also run from g1t's open-source pool, after the same card check.
216	Agents never run from the pool. Workflow jobs with `runs-on: self-hosted`
217	run on the workspace's own machines (`g1t-runner`, any OS) at $0, on every
218	plan; a workspace can send agent work there too. Each workspace decides how its agents
219	reach a model: its own provider (connected under Integrations, billed by
220	the provider) or g1t's hosted models (while payments are in test mode,
221	only for a few invited workspaces; a trial does not open them, and an
222	agent assigned without a model is refused with `no_model`); the sandbox is g1t's unless the work goes to
223	the workspace's own runners.
224	When the plan refuses a start, these calls answer with a failure whose
225	message says what to do and where (such as `/acme/-/billing`). When every
226	agent slot of the workspace is busy, the message starts "Waiting for a
227	free slot" and the work starts by itself when one finishes.
228
229	- Hand off an outcome: `POST {repo}/plans` with `brief`: what should be
230	true when the work is done. A planner reads the repository and proposes
231	issues, each with what done means (`done`), the files it touches, and what it depends
232	on. Read it with `GET {repo}/plans/{plan}` until `status` is `ready`
233	(a minute or two), then `POST {repo}/plans/{plan}/apply` with
234	`{"assign": true}`. Agents start at once on every issue that depends on
235	nothing and on the rest as what they depend on lands. `keep` opens only
236	some of the issues, by position counting from 1.
237	- Assign one issue: `POST {repo}/issues/{number}/assign`. The agent
238	opens a pull request, waits for the repository's workflows on it and is
239	sent back with the failing jobs' log tails when one fails, is reviewed by
240	a second agent, revises, and catches up when main moves. After its
241	revisions (`max_revisions`, 2 by default) only a required check still
242	failing holds it for a person. There is no model or
243	agent count to choose: to put more agents to work, assign more issues.
244	On g1t's hosted models, Auto routes each job to the cheapest of three
245	tiers that can do it, fast, standard and most capable (the model behind
246	each is today's, and moves to newer models as g1t adopts them; a
247	retired model is never used): catching up, answering, planning and
248	reviews of small changes that touch no sensitive path start fast;
249	making changes, revising and other reviews start standard; reviews of
250	very large changes and issues labelled
251	`architecture` start most capable. A failed attempt moves the next one
252	up a tier (two in a row: most capable), and a repository's own recent
253	runs move work down or up. Each run states its model and why in one
254	line, on the run and in the session. An owner can pin a tier per kind of
255	work instead (`PUT /workspaces/{workspace}/model-routes`, `model`
256	`small`, `large` or `frontier` with `connection_id` null).
257	- Who a g1t pull request is for: g1t is the `author` (`username`
258	`g1t`, `kind` `agent`) of every pull request it makes and every issue it
259	files at work; `requested_by` is the person who asked (null when nobody
260	did, as for a security update). That person answers for it as an author
261	would: they may update, close and steer it without Triage, cannot
262	approve it, are never asked to review it, see it in their own lists, and
263	its sandboxes, workflows and previews are trusted as they are. Webhooks
264	carry `data.author` and `data.requested_by`; Actions payloads
265	`pull_request.user` (a `Bot`) and `pull_request.requested_by`.
266	- Put an agent on something in one step: `POST {repo}/issues/delegate`
267	with `title` and `body` (what to do, in plain words, and what done
268	means if you know it). It opens the issue and assigns g1t at once; it needs the
269	Write role, and nothing opens without it. The issue opens even when the
270	agent cannot start: `agent.status` is `started` (with `pull`), `queued`,
271	or `not_started` with `agent.code` (`not_paid`, `trial_used`, `limit`,
272	`paused`, `issue_cap`, `billing_unavailable`, `no_model`), `agent.message`
273	and `agent.fix_url`.
274	- How sure g1t is of a change: once g1t finishes, the pull
275	request's `confidence` is `high`, `medium` or `low` with short `reasons`
276	("tests not added", "3 revisions"), from its required checks, revisions, review,
277	tests, size, reach, guardrails and unanswered questions. The agent's own
278	word (`self_reported`, `uncertain_about`) can only lower it. With the
279	repository setting `hold_low_confidence` on (the default), a change rated
280	low waits for a person's approval instead of merging by itself.
281	- Steer a working agent: `POST {repo}/pulls/{number}/messages` with
282	`body`. It reads the message at its next step.
283	- g1t's runs talk to each other. g1t asks the agent on another
284	pull request a question, or hands it work, with `agent` `message`
285	(`kind` `question` or `handoff`, and `from_number`, its own pull
286	request). The other agent replies with `agent` `answer`
287	(`POST {repo}/messages/{id}/answer`); one that is not at work is woken
288	to answer, in its own pull request's sandbox. Plans show these exchanges under
289	"Agents talking". From any other caller, `agent` `message` sends a plain
290	message.
291
292	Every one of these is also on the MCP server. Its tools are resources,
293	each with an `action`: `search`, `repository`, `issue`, `pull_request`,
294	`agent`, `plan`, `memory`, `workflow`, `secret`, `webhook`, `access`,
295	`workspace`, `notifications` and `account`. Call `tools/call` with the tool's name and
296	`arguments` holding `action` and its inputs, such as
297	`{"name": "issue", "arguments": {"action": "get", "repo": "acme/web", "number": 12}}`.
298	The flow above is: `issue` `get`, `memory` `recall`, `pull_request`
299	`create` (with `issue`), push, `pull_request` `record_session` as you go,
300	`pull_request` `ready` with `summary`; read `overlaps`, `behind`,
301	`statuses` and `required_checks` on `pull_request` `get`, and
302	`repository` `check_names` for the names a branch can require;
303	`workflow` `list_check_runs` and `check_run_annotations` for what a
304	commit's checks say. `agent` `delegate` and `agent` `assign` hand work
305	to g1t's agent; `plan` `create`, `get` and `apply` plan an outcome;
306	`memory` `remember` saves a fact. `notifications` reads and answers the
307	inbox of the person a token acts for: `list` (unread threads, each with a
308	`reason` such as `agent` or `review_requested`), `done`, `subscribe`,
309	`watch` and more; g1t's own token cannot use it. `search` runs `code`,
310	`notifications` runs `list` and `account` runs `whoami` when `action` is
311	left out. A call missing a required field says
312	which, such as "issue.get needs number.". The earlier one-tool-per-operation
313	names (`get_issue`, `create_pull_request`, …) still answer for now but are
314	no longer listed. Every tool and action, with its required fields and
315	scope: https://docs.g1t.sh/reference/mcp/
316
317	g1t's own token can never change a repository's details, rename it
318	or its branches, make it public or private, archive, transfer, delete,
319	restore or purge it, or delete a workspace. MCP tools take the repository
320	as `repo`, written `owner/name`.
321
322	## Scopes
323
324	Every access token and OAuth sign-in has scopes, `resource:level`:
325	`repo`, `code`, `issues`, `pull_requests`, `workflows`, `checks`, `deployments`, `memory`,
326	`account`, `notifications`, `access`, `webhooks`, `secrets`, `runners`, `models` (read, write or admin
327	as each has them), `agents:run`, `workspace:read` and `workspace:admin`. A higher
328	level includes the lower. A token may expire. It reaches every workspace
329	and repository its owner can (a workspace's token, that workspace only);
330	what a call may do is the owner's role and the token's scopes together.
331	A token sees only the MCP tools and actions its scopes allow. A missing scope answers `403` with
332	`{"error": {"code": "forbidden", "message": "This access token needs the issues:write scope to use create_issue.", "needed_scope": "issues:write"}}`.
333	Pushing needs `code:write`; cloning a private repository `code:read`.
334	Pushing commits that add, change or delete files under `.g1t/workflows/`
335	or `.github/workflows/` also needs `workflow_files:write` (in no preset
336	but full access); a workflow job's token never has it. A fine-grained
337	token reaches one workspace (or only its owner's account), all, chosen or
338	only public repositories of it, with permissions (`contents`, `issues`,
339	`workflows`, …) mapped to these scopes; a workspace may require an owner to approve one first. For
340	an agent, use the Agent preset (every read scope but `runners:read`, plus `code:write`,
341	`issues:write`, `pull_requests:write`, `agents:run`, `memory:write`,
342	`notifications:write`). For CI, the CI preset (`repo:read`, `code:read`,
343	`code:write`, `packages:read`, `packages:write`, `workflows:read`,
344	`workflows:write`, `deployments:read`, `deployments:write`). `models:write` (sending requests through the AI
345	Gateway, which spends AI credit) is in no preset but full access.
346	OAuth clients may send `scope`;
347	the person can untick any; asking for none gives the Agent preset. Tokens
348	from device sign-in (above) have full access. Guide:
349	https://docs.g1t.sh/guides/authentication/#scopes
350
351	## AI Gateway
352
353	Your own code can call models through g1t in either format, with a
354	workspace access token with `models:write` as the API key (`x-api-key` or
355	`Authorization: Bearer`):
356	- Anthropic's Messages API at `https://models.g1t.sh/anthropic`:
357	`POST /v1/messages` (streamed or not), `POST /v1/messages/count_tokens`.
358	- OpenAI's at `https://models.g1t.sh/openai/v1`: `POST /chat/completions`
359	(streamed or not, function tools, `response_format`, `reasoning_effort`),
360	`POST /embeddings`, `GET /models` (what the workspace can use, with g1t's
361	`pricing` per million tokens and `billed_to`).
362	Any model works in either format; the proxy translates. Model ids:
363	`anthropic/claude-haiku-5-5` (cheapest Claude; priced higher above 100,000
364	prompt tokens), `anthropic/claude-sonnet-5-5`, `anthropic/claude-opus-5-5`,
365	`anthropic/claude-haiku-4-5` (bare Claude ids work too), and open models on
366	Workers AI such as `workers-ai/@cf/openai/gpt-oss-120b`,
367	`workers-ai/@cf/zai-org/glm-5.3-flash`, and embeddings
368	`workers-ai/@cf/baai/bge-m3`. Requests on g1t's models are charged at the
369	model's list price (no markup in beta) from included usage and AI credit,
370	never the agent rate. The workspace's own providers under Integrations (an
371	Anthropic or OpenAI key, any OpenAI- or Anthropic-compatible endpoint such
372	as vLLM or Ollama) take the models listed in their `config.gateway_models`
373	(ids or `prefix`; `ns/` strips `ns/`; Anthropic keys default to
374	`claude-*`), come first, and are free on g1t, only counted. Set them with
375	`connect_integration` or `update_integration` (`workspace:admin`); keys are
376	write-only. Refusals are in the route's format: `402` (`billing_error` /
377	`insufficient_quota`) when out of AI credit, over the spend limit or not on
378	the plan; `403` without `models:write` or with a personal token; `404` for a
379	model no provider offers; `400` on g1t's models for fast mode (`speed`),
380	`inference_geo`, `fallbacks`, `container`, server tools or non-function
381	tools, `web_search_options` (all fine on the workspace's own provider). For
382	Claude Code: `ANTHROPIC_BASE_URL=https://models.g1t.sh/anthropic` and
383	`ANTHROPIC_AUTH_TOKEN=g1t_…`. The log (30 days, no prompts; each request's
384	`format`, `provider`, `connection`, model and tokens):
385	`GET /workspaces/{workspace}/gateway/requests` or the `billing` tool's
386	`gateway_requests` action, with `models:read`. Guide:
387	https://docs.g1t.sh/guides/ai-gateway/
388
389	## Access and roles
390
391	Everyone's access to a repository is a role: `read` (read, clone, open
392	issues and pull requests, comment), `triage` (also label, assign, close),
393	`write` (also push, merge, and put agents to work: anything that spends
394	compute), `maintain` (also settings, branch protection, guardrails) or
395	`admin` (also webhooks, secrets, deployments, domains, who has access,
396	rename, archive, visibility, default branch). Owners of a workspace have
397	admin on all of its repositories and alone transfer or delete them;
398	members get the workspace's base permission (write unless owners change
399	it); anyone can be given a role on one repository, as an outside
400	collaborator; anyone reads a public repository. The highest wins. A
401	private repository you cannot read answers `404`; one you can read but
402	lack the role for answers `403` naming the role needed. An agent works
403	with the role of the person it acts for on its repository, never more
404	than `write`, and its token can never change who has access. An outside
405	collaborator with `write` can put agents to work; the runs are charged to
406	the repository's workspace, and their agents are told the project's memory,
407	never the workspace's. Who can do what elsewhere: deployments are seen with
408	`read` (on a public repository, by anyone, build logs included), deployed
409	with `write`, configured (settings, domains) with `admin`; project settings
410	and dependencies need `maintain`; repository webhooks, secrets and
411	variables need `admin`, seeing them included; security alerts need
412	`write` (dismissing a dependency alert too), dismissing or reopening a
413	secret alert `admin`, turning security updates on or off `maintain`;
414	enabling or disabling a workflow needs `maintain`; plans and project memory
415	are read by anyone who can read the repository, and project memory is
416	changed with `write`; workspace memory is for members. People: the
417	workspace's Settings → Members (`g1t.sh/<owner>/-/people`, with an Outside
418	collaborators tab and the Base permission for owners); a repository's
419	Settings → Access (`g1t.sh/<owner>/<repo>/settings/access`); invitations
420	are answered at `g1t.sh/<owner>/<repo>/invitations`.
421	`GET {repo}/collaborators/{username}/permission` gives a role and what it
422	allows. Guide: https://docs.g1t.sh/guides/access-and-roles/
423
424	Teams group a workspace's members (`GET /workspaces/{workspace}/teams`;
425	the `team` MCP tool). A team given a role on a repository gives it to
426	everyone in it and its child teams, and `source` is then `team`.
427	`@workspace/team` in a comment tells the team's people; a pull request can
428	ask a team to review (`POST {repo}/pulls/{number}/requested_reviewers` with
429	`team_reviewers`), and the team may pick who. Guide:
430	https://docs.g1t.sh/guides/teams/
431
432	A CODEOWNERS file (`.g1t/CODEOWNERS`, `.github/CODEOWNERS`, `CODEOWNERS`,
433	`docs/CODEOWNERS` or `.gitlab/CODEOWNERS`, the first found on the default
434	branch) asks owners to review pull requests that change their files. With
435	`require_code_owner_review` on (`PATCH {repo}/settings`), a merge waits for
436	their approval; `code_owners` on `GET {repo}/pulls/{number}` says whose is
437	missing, and `GET {repo}/codeowners/errors` checks the file. Guide:
438	https://docs.g1t.sh/guides/codeowners/
439
440	## Manage a repository
441
442	People with the admin role rename it (`POST {repo}/rename` with `name`; the
443	old address redirects), make it public or private
444	(`POST {repo}/visibility` with `private` and its full name in `confirm`),
445	archive or unarchive it (`POST {repo}/archive`, `POST {repo}/unarchive`),
446	and owners of its workspace delete it (`DELETE {repo}` with its full name
447	in `confirm`). A deleted
448	repository can be restored for 30 days (`POST {repo}/restore`, listed by
449	`GET /workspaces/{workspace}/repos/deleted`) and is then purged; its name
450	stays taken until then, or until `POST {repo}/purge`. Maintain changes the
451	description, `website` and `topics` with `PATCH {repo}`, admin the
452	`default_branch`, and write renames branches with
453	`POST {repo}/branches/{branch}/rename` and `new_name` (slashes in the
454	branch URL-encoded); only admin renames the default branch. An archived
455	repository is read-only: pushes and merges are refused, issues and pull
456	requests are locked, and agents and workflows do not run on it.
457
458	## Integrations
459
460	A workspace's owners connect it to outside systems on its Integrations
461	page, or with `POST /workspaces/{workspace}/integrations`:
462
463	- Its own model providers (`anthropic`, `openai`, `gemini`, `xai`,
464	`mistral`, `deepseek`, `azure_openai`, `openrouter`, `groq`, `together`,
465	`fireworks`, `cerebras`, `anthropic_endpoint`, `openai_endpoint`), as many
466	as it uses, with each
467	kind of work routed to one of them or to g1t's hosted models
468	(`PUT /workspaces/{workspace}/model-routes`). Those providers bill the
469	workspace; g1t charges only each run's sandbox time, at cost plus 20%
470	(nothing on the workspace's own runners). Sandboxes never hold a key.
471	- Alerts (`sentry`, `datadog`, `webhook`): each problem opens one issue
472	in a chosen repository, optionally with an agent put on it at once.
473	Senders sign requests to `https://api.g1t.sh/hooks/{integration}`.
474	- Trackers (`jira`, `linear`): `GET {repo}/context?reference=TECH-1234`
475	fetches a ticket; `POST {repo}/issues/import` with `reference` (and
476	`assign`) opens a linked issue. Agents get tickets their work mentions in
477	their starting context. Ticket text is reference material, never
478	instructions.
479
480	## Webhooks
481
482	`POST {repo}/hooks` (or `/workspaces/{workspace}/hooks` for every
483	repository in a workspace) with `url` and optional `events` sends events
484	to that HTTPS address as they happen, signed in `X-G1t-Signature-256`
485	(HMAC-SHA256 of the body), retried for about seven hours. Deliveries,
486	with request and response, are at `…/hooks/{id}/deliveries`.
487
488	## GitHub Actions
489
490	GitHub Actions workflows run on g1t unchanged, from `.g1t/workflows/`
491	(g1t never reads `.github`): moving a repository is `git mv .github .g1t`.
492	Runs, jobs and logs are at GitHub's own routes under
493	`{repo}/actions/...`. A run on a pull request's head is a check: pending
494	holds the merge, failure refuses it and sends g1t back to fix it.
495	Checks: CI and integrations report on commits with a token holding
496	`checks:write` and the Write role: statuses
497	(`POST {repo}/statuses/{sha}` with `state` pending/success/failure/error,
498	`context`, `description`, `target_url`; `GET {repo}/commits/{ref}/status`
499	combines them) and check runs (`POST {repo}/check-runs` with `name`,
500	`head_sha`, `status`, `conclusion`, `output` {`title`, `summary`,
501	`text`, `annotations`}, `actions`; `PATCH {repo}/check-runs/{id}` to
502	complete one). A required check is met by a status or a check run of its
503	name. g1t's agents read checks and never report them. Guide:
504	https://docs.g1t.sh/guides/checks/
505	Secrets and variables are one list per repository (site:
506	`g1t.sh/<owner>/<repo>/settings/secrets`) and per workspace: each row is a
507	key, Secret or Config, the environments it applies to (all, or e.g.
508	production/preview, or a job's `environment:`), and whether workflows,
509	deployments or both read it. API: `{repo}/actions/secrets` and
510	`{repo}/actions/variables` (GitHub's routes) with extra `environments`,
511	`available_to`, `repositories`, `note`, `id`. Every job gets
512	`secrets.G1T_TOKEN` (`GITHUB_TOKEN` is its alias): a token for that job
513	only, reaching its repository only, with the scopes its `permissions:`
514	give (default `contents: read`, `packages: read`), revoked when the job
515	ends; what it changes starts no workflows except `workflow_dispatch` and
516	`repository_dispatch` (`POST {repo}/dispatches`). It cannot change
517	secrets. A job naming an environment with protection rules (required
518	reviewers, wait timer, branch limits; `PUT {repo}/environments/{name}`)
519	waits until they pass, and reviewers approve it at
520	`POST {repo}/actions/runs/{id}/pending_deployments`. A pull request's
521	runs from outside may wait as `action_required` until someone with Write
522	approves them (`POST {repo}/actions/runs/{id}/approve`); agents cannot
523	approve runs or deployments. A pull request's runs and preview are trusted
524	only when its author has `write` or higher on the repository (a member or
525	an outside collaborator), or is g1t working on its own; for one g1t made,
526	the role of whoever asked for it (`requested_by`) counts. Anyone else's run
527	with config only. Guide:
528	https://docs.g1t.sh/guides/secrets-and-variables/
529
530	Self-hosted runners: a job with `runs-on: self-hosted` (or
531	`[self-hosted, linux, gpu]`, or `{group: name}`) waits until a runner of the
532	workspace's with every label takes it. Owners add one under
533	`g1t.sh/<workspace>/-/runners`: a one-hour registration token, then
534	`g1t-runner register --url https://g1t.sh --token g1trt_…` and
535	`g1t-runner run`. Runners only connect out. API:
536	`/workspaces/{workspace}/actions/runners`, `.../runner-groups`,
537	`.../runner-settings` (and the same under `{repo}/actions/` for a
538	repository's own); MCP: the `workflow` tool's `list_runners`,
539	`create_runner_token`, `remove_runner`, runner group and settings actions
540	(`runners:read`/`runners:admin`). Pull requests from forks never run on them
541	unless allowed. Guide: https://docs.g1t.sh/guides/self-hosted-runners/
542
543	## Projects
544
545	A project is what a workspace builds and runs; every repository is a
546	project of its own name (`g1t.sh/<owner>/<project>` opens its overview; its
547	code is under `/code`; every repository address still works). Deployments,
548	secrets and variables belong to the project; branches, pull requests,
549	review and merge rules to its repository (Settings → Repository). Guide:
550	https://docs.g1t.sh/guides/projects/
551
552	## Security
553
554	A push that adds a known key or token format (AWS, GitHub, GitLab, Stripe
555	live, Slack tokens and webhooks, Google, Anthropic, OpenAI, npm, g1t,
556	SendGrid, PEM private keys, service-role JWTs) is refused with every secret
557	listed by `file:line` in git's output and a link to allow it; this includes
558	an agent's push to its pull request. A very large push is scanned after it
559	lands, not before: it goes through, its new commits (any branch) are
560	scanned in the background, and a secret found is an open alert, emailed to
561	the workspace's owners when it looks real. History is scanned once in the
562	background. Never commit a secret: read it from the environment. Values are
563	judged by the value alone, never the file's path: a documented example key,
564	a value containing example/sample/dummy/fake/placeholder/changeme/notreal/
565	redacted/xxxxx, one that counts up (six or more, like 123456), one
566	character five or more times, a repeated short piece, or too little
567	randomness is a "likely test value": listed apart, never blocks a push,
568	never counted critical. Any other fixture carries `g1t:allow-secret` in a
569	comment on its line. Alerts are `open`, `dismissed` or `fixed`. Dismissing a
570	secret alert needs `admin`, with a reason (`false_positive`,
571	`used_in_tests`, `wont_fix`: dismissed, and pushes carrying it go through;
572	`revoked`: fixed) and an optional comment (500 characters); a dependency
573	alert needs `write` (`fix_started`, `no_bandwidth`, `tolerable_risk`,
574	`inaccurate`, `not_used`). Reopen undoes it. API:
575	`GET {repo}/security/alerts` (`state`, `kind`),
576	`POST {repo}/security/alerts/{id}/dismiss` (`reason`, `comment`),
577	`POST {repo}/security/alerts/{id}/reopen`; MCP `repository` actions
578	`security_alerts`, `dismiss_alert`, `reopen_alert` (`repo:read` to list,
579	`repo:admin` to dismiss or reopen). Lockfiles (npm, pnpm, yarn, Cargo, Go,
580	Python) on the default branch are checked against OSV on every push to it
581	and daily. Security updates (on by default; `maintain` turns them off): for
582	each vulnerable dependency with a fix, g1t itself opens a pull request
583	authored by `g1t` from `g1t/security/<package>-<version>`, raising the
584	version in each lockfile with the ecosystem's own tool; it merges through
585	the branch's required checks and merge queue. A newer update for the same
586	package, or the package no longer being vulnerable, closes it as
587	superseded. Only when the bump fails, or required checks fail because code
588	must change, does g1t open an issue and assign it to g1t (the session
589	shows "started by g1t"). With no fix published, the alert links the
590	advisory and is checked again daily. `.g1t/dependencies.yml` (version
591	updates) is read and validated, but no version update pull requests are
592	opened yet. `g1t` is not an account and cannot be signed in to.
593	Guide: https://docs.g1t.sh/guides/security/
594
595	The security suite (scopes `security:read`, `security:write`; MCP tool
596	`security`). A push refused for a secret links to its alert, where anyone
597	with `write` bypasses it with a reason (`false_positive`, `used_in_tests`,
598	`will_fix_later`), or asks owners when the workspace delegates bypasses;
599	an agent never bypasses: take the secret out and push again. Code
600	scanning: upload SARIF 2.1.0 to `POST {repo}/code-scanning/sarifs`
601	(`commit_sha`, `ref`, `sarif` gzipped then base64); default-branch results
602	become alerts, pull request results (`refs/pull/<n>/head`) become line
603	comments and the `Code scanning` status. `Dependency review` is a status on
604	every pull request that changes a lockfile. Fix what either reports in your
605	own pull request; both can be required checks. Also: custom patterns,
606	validity checks, `GET {repo}/dependency-graph/sbom` (SPDX 2.3, in `sbom`),
607	`GET {repo}/dependency-graph/compare/{base...head}`, and a workspace
608	overview. On private repositories these need the Security and quality
609	activation (`402` without it); public repositories have them free.
610	Guides: https://docs.g1t.sh/guides/security/secret-protection/,
611	https://docs.g1t.sh/guides/security/code-scanning/,
612	https://docs.g1t.sh/guides/security/supply-chain/
613
614	## Deployments
615
616	Part of the g1t plan ($20 a month per workspace, started by an owner
617	under the workspace's Billing). No quotas: unlimited projects and
618	previews (never charged); builds by the second, requests ($0.36/M), CPU
619	($0.024/M ms) and custom domains ($0.12 a month each) metered from the
620	first at cost + 20%, from the plan's $10 first. The trial
621	never covers deployments. Then someone with admin on the repository
622	turns deployments on for a project (Settings → Deployments, or Deploy on
623	its overview). Production deploys from the default branch to
624	`https://<project>-<owner>.g1t.page` on each push; every branch with an
625	open pull request gets a preview at
626	`https://<project>-git-<branch>-<owner>.g1t.page` (a fork's pull request is
627	`pr-<n>`), shown on it as the check `g1t / deploy` (`g1t / deploy (<project>)`
628	for a workspace's other projects). Builds and running apps
629	read the project's secrets and variables available to Deployments, each
630	key's Production or Preview row. Workers projects (`wrangler.jsonc`) and
631	static sites build without configuration. Previews come down when the pull
632	request closes and after idle days.
633	Guide: https://docs.g1t.sh/guides/deployments/
634
635	A repository's deployments, wherever they run, are one list: `source` is
636	`api` (reported from any CI), `actions` (a g1t Actions job with
637	`environment:`) or `g1t_page` (g1t.page builds, ids `dpl_…`, environments
638	`production` and `preview`). Read with `GET {repo}/deployments` (filters
639	`environment`, `ref`, `sha`, `task`, `state`, `source`, `creator`, `page`,
640	`per_page`), `GET {repo}/deployments/{id}`, `GET {repo}/deployments/{id}/statuses`,
641	`GET {repo}/environments` and `GET {repo}/environments/{environment}`
642	(`deployments:read`, Read role). Report from a CI with
643	`POST {repo}/deployments` (`ref`, optional `environment` (default
644	`production`), `sha`, `task`, `description`, `payload`, `state`,
645	`environment_url`, `log_url`), then
646	`POST {repo}/deployments/{id}/statuses` with `state` (`queued`,
647	`in_progress`, `success`, `failure`, `error`, `inactive`) and
648	`environment_url` (`deployments:write`, Write role). A success makes the
649	environment's older successful deployments `inactive` unless
650	`auto_inactive: false`. Each status sets the commit check
651	`deploy / <environment>`, which a ruleset's `required_deployments` rule can
652	require. A g1t Actions job with `environment:` (or `{name, url}`) reports
653	by itself, one deployment per run and environment; `deployment: false`
654	opts out. MCP: the `workflow` tool's `list_deployments`, `get_deployment`,
655	`create_deployment`, `deployment_statuses`, `create_deployment_status`,
656	`list_environments`, `get_environment`. Webhooks: `deployment.created`,
657	`deployment_status.created`. Guide: https://docs.g1t.sh/guides/deployments-api/
658
659	## Search
660
661	`GET https://api.g1t.sh/search?q=<query>&type=<type>` (MCP: `search`, action `code`, the default)
662	searches all of g1t: repositories (name, description, topics, README), code
663	on default branches, issues, pull requests, people and workspaces. No token
664	needed for public results; with one, private results the token's person
665	can read are included (their workspaces' repositories, and those they were
666	given a role on), checked against current membership and roles. `type` is
667	`repositories`, `code`, `issues`, `pulls` or `people` (worked out from the
668	qualifiers when left out); `page` and `per_page` (at most 50) page through.
669	The query takes words, `"exact phrases"`, `-word` to leave out, and
670	`repo:owner/name`, `org:<workspace>`, `language:<lang>`, `path:<prefix or
671	*.glob>`, `is:issue`, `is:pr`, `is:open`, `is:closed`, `is:merged`,
672	`author:<username>`, `label:<label>`. Code search matches any run of three
673	characters or more; vendored directories, lockfiles, binaries and files
674	over 512 KB are not indexed. Results give `counts` per type and each hit's
675	`snippet` or code `lines` as parts with `highlight`. On the site:
676	`https://g1t.sh/search?q=`, ⌘K, and `https://g1t.sh/explore` for public
677	projects by activity, language (`?language=`) and topic (`?topic=`).
678	The `search` tool's `context` action stays the search of one workspace's
679	context hub. Guide:
680	https://docs.g1t.sh/guides/search/
681
682	## Facts
683
684	- API base: `https://api.g1t.sh`. `GET /` lists the main URLs as templates;
685	every operation is in the OpenAPI document.
686	Auth: `Authorization: Bearer g1t_…`. Public data needs no token. Errors are
687	`{"error": {"code": "…", "message": "…"}}` with codes `unauthenticated`
688	(401), `payment_required` (402, when the plan or a limit refuses compute),
689	`forbidden` (403, with `needed_scope` when the token lacks a
690	scope), `not_found` (404), `conflict` (409), `invalid` (422),
691	`rate_limited` (429, with `Retry-After`: 1,000 requests a minute per
692	token, 60 per IP address without one; wait, then retry). Each
693	operation's scope is `x-scope` in the OpenAPI document. The full description is at https://api.g1t.sh/openapi.json.
694	- Git remote: `https://g1t.sh/{workspace}/{repo}.git`. In API paths,
695	`{owner}` is the workspace. Pull request forks:
696	`https://g1t.sh/pulls/{pull_request_id}.git`. SSH is not available.
697	- Limits: 1 GB per repository, 32 MB per file, 100 MB per push. Every
698	current limit, why it exists and the workaround:
699	https://docs.g1t.sh/about/limitations/
700	- Forgotten password: https://g1t.sh/forgot (the person does this, in a
701	browser).
702	- Times are RFC 3339 in UTC.
703	- OAuth 2.1 for applications: metadata at
704	`https://api.g1t.sh/.well-known/oauth-authorization-server`; authorization
705	code with PKCE (S256), public clients, dynamic registration.
706	- A pull request whose required checks have not passed (failed, still
707	running, or not reported yet) is refused a merge with `409`, saying
708	which; where the repository allows bypassing them
709	(`allow_ignoring_checks`), someone who can merge can send
710	`{"ignore_checks": true}`. Checks that are not required never hold a
711	merge. With the merge queue on, the workflows behind required checks
712	need `merge_group` in their `on:`.
713	- Landing fast-forwards the default branch: g1t makes no merge commit on
714	main. Bringing a pull request up to date makes a merge commit on its own
715	branch.
716	- Pricing (https://g1t.sh/pricing): the forge is free. One plan, g1t, at
717	$20 a month per workspace with unlimited members, includes $10 of usage
718	at cost + 20% (unused does not roll over). Compute needs the plan or a
719	card check (never charged); the $5 trial needs a credit or debit card,
720	not a prepaid one. No quotas on the plan: everything is metered from the
721	first unit at cost + 20%, and only the spend limit stops work. Every
722	workspace has 1 GB of private storage and 50,000 git operations a month
723	free; past them, the plan pays ($0.60/GB-month, $0.18/1,000) and a free
724	workspace is held (pushes to private repositories stop; git slowed to 60
725	an hour). Limits: $100 in a
726	paid workspace's first month, rising as payments clear; owners set a
727	spend limit, prepay, or ask with Raise my limit. Caps: $2 a run and $10
728	an issue by default. A spend spike pauses new compute until an owner
729	chooses Keep going or Stop (`/<workspace>/-/billing`). Audit log: 90 days
730	on every plan.
731
732	## More
733
734	- [Quickstart](https://docs.g1t.sh/quickstart/)
735	- [How g1t works](https://docs.g1t.sh/concepts/overview/)
736	- [Search and Explore](https://docs.g1t.sh/guides/search/)
737	- [g1t's agent](https://docs.g1t.sh/guides/working-with-g1t/)
738	- [Outcomes and plans](https://docs.g1t.sh/guides/outcomes/)
739	- [Talking to agents](https://docs.g1t.sh/guides/talking-to-agents/)
740	- [Bring your own agent](https://docs.g1t.sh/guides/bring-your-own-agent/)
741	- [The merge queue](https://docs.g1t.sh/guides/merge-queue/)
742	- [Sessions and why-blame](https://docs.g1t.sh/guides/why-blame/)
743	- [Forks and branches](https://docs.g1t.sh/concepts/forks/)
744	- [Accounts and sign-in](https://docs.g1t.sh/guides/authentication/)
745	- [Workspaces and tokens](https://docs.g1t.sh/guides/workspaces/)
746	- [Access and roles](https://docs.g1t.sh/guides/access-and-roles/)
747	- [Teams](https://docs.g1t.sh/guides/teams/)
748	- [CODEOWNERS](https://docs.g1t.sh/guides/codeowners/)
749	- [Managing a repository](https://docs.g1t.sh/guides/managing-repositories/)
750	- [Integrations](https://docs.g1t.sh/guides/integrations/)
751	- [Model providers](https://docs.g1t.sh/guides/models/)
752	- [AI Gateway](https://docs.g1t.sh/guides/ai-gateway/)
753	- [Webhooks](https://docs.g1t.sh/guides/webhooks/)
754	- [GitHub Actions](https://docs.g1t.sh/guides/actions/)
755	- [Self-hosted runners](https://docs.g1t.sh/guides/self-hosted-runners/)
756	- [Usage and billing](https://docs.g1t.sh/guides/usage-and-billing/)
757	- [Git](https://docs.g1t.sh/guides/git/)
758	- [MCP tools](https://docs.g1t.sh/reference/mcp/)
759	- [API reference](https://docs.g1t.sh/reference/api/)
760	- [Rate limits](https://docs.g1t.sh/reference/rate-limits/)
761	- [What g1t can't do yet](https://docs.g1t.sh/about/limitations/)
762	- [An open letter to Cloudflare](https://docs.g1t.sh/about/open-letter-to-cloudflare/)
763	- [Source](https://g1t.sh/flagon-io/g1t), MIT licensed
764
765	## Help, status and policies
766
767	- [Status](https://status.g1t.sh/): whether each part of g1t is working now, 90 days of uptime, and incidents; the same as JSON at https://status.g1t.sh/status.json
768	- [Support](https://g1t.sh/support): where to get help (hey@flagon.io)
769	- [Security](https://g1t.sh/security): how g1t protects code and accounts, and responsible disclosure (hey@flagon.io, https://g1t.sh/.well-known/security.txt)
770	- [Policies](https://g1t.sh/policies): [Terms of Service](https://g1t.sh/policies/terms), [Privacy Policy](https://g1t.sh/policies/privacy), [Acceptable Use](https://g1t.sh/policies/acceptable-use), [Refunds and Cancellation](https://g1t.sh/policies/refunds), [Subprocessors](https://g1t.sh/policies/subprocessors)
771	- g1t is made by Flagon, Inc. (https://www.flagon.io)