Skip to content
Sign up

flagon-io/g1t

public
00

Where people and agents ship software together. The open-source git platform for the whole job: issues, agents, checks and deploys to the edge.

gitagentscloudflarerustforgemcp
2branchesTags
syntaqxBlame is kept by commit and path, reads ahead and stops in time, and crawlers are asked to leave it alone; a push works out its repository's About, and a run's reads are kept for the next. Workers Logs at full sampling (2026-10-14, ~21:30-22:15 UTC) had blame as 56 of 101 repos answers over a second (median 2.4 s, up to 20.3 s), and the site's own logs showed who asked: 1,154 blame pages in four hours, about 1,015 of them ClaudeBot, 93 Googlebot, the rest Amazonbot, GPTBot and Reflectionbot, and a handful from people, each a file at a commit (`/flagon-io/g1t/blob/<sha>/<path>?blame=1`, `/pulls/<id>/blob/...`). Nothing was kept, so every one walked up to 400 commits again, reading the trees to the file one commit at a time, each read a round trip. Now repos keeps a blame in the colo's cache by repository, commit and path (blame.rs `cache_url`, for good; one cut short for ten minutes), finding the commit first when asked by branch; the walk reads the trees of the next 32 commits level by level at once, then at once the texts it will compare, reads each tree and text once, stops as soon as every line has its commit, and stops after 6 s (BLAME_WALK_MS) or 2,500 reads, giving what is left to the oldest commit it reached them in, marked partial as a history past 400 commits already was. robots.txt disallows `/*?blame=` and `/*&blame=` (app/lib/robots.ts), the Blame links are `nofollow`, and a blame page says `noindex`; the file itself stays open to crawlers. The About's 22.7 s and 22.0 s "about" invocations were its background run hitting its 22 s budget in the answer's waitUntil (the answer itself was quick): its history was read 200 commits a page, and each page after the first starts wherever the last ended, which moves with every push, so g1t's 728 first-parent commits were four reads with three of them read whole from the store again each time; pages are now 1,000, the length lib.rs already reads (MAX_ANCESTRY), so most repositories are one spliced read. And refresh_later never waited for the cache writes its run started (store.rs keeps trees, sizes and logs through Deferred, which only push handling settled), so a run could leave the next to read the same again; it now settles them. A push that moves the default branch works the About out after recording the push, within what is left of 24 s, so a view after a push finds it current; a run that is slow or partial logs `about_run` with `took_ms`. last_commits no longer waits on its two writes (the memo, which also goes to KV, and the finished answer) before answering. Blob had no waste worth taking out: its slow answers are the store's first read of a file, which is then kept. Repos and work log a quarter of requests (head_sampling_rate 0.25), down from every one for the stall hunt (no ~60 s stall recurred); slow RPC answers still log `answered_ms`. The deploy guide's Logs section, why-blame and managing repositories say so. Checked: cargo test -p g1t-repos (356 passing, 11 new), cargo test -p g1t-api openapi (9), npm test in apps/web (944 passing, 2 new), npx tsc -b --force, npm run test:deploy (92), node scripts/deploy.mjs manifest --check, and the docs build.

README.md

g1t

Git tracks changes. g1t coordinates change.

g1t is a git platform built for teams whose code is written by people and their coding agents at the same time. You keep your own harness (Claude Code, Codex, Cursor, Copilot or anything that speaks MCP); g1t is the place they all work through. It hosts the repositories, shows who and what is working on each project right now, keeps parallel work from colliding, tests changes together before they land, deploys them, and remembers why each change was made.

g1t is MIT-licensed and in alpha. Use it on g1t.sh (anyone can sign up), or run it yourself. It runs on Cloudflare Workers, with repositories in Cloudflare Artifacts, and the core runs in Docker.

Try it

# 1. Sign up and make a workspace at https://g1t.sh/register.

# 2. Install the g1t CLI (Windows: irm https://g1t.sh/install.ps1 | iex).
curl -fsSL https://g1t.sh/install | sh
g1t auth login

# 3. Install g1t's Claude Code plugin. It adds g1t's MCP server and records
#    your sessions; run /mcp in Claude Code and sign plugin:g1t:g1t in.
claude plugin marketplace add https://g1t.sh/flagon-io/g1t-plugins.git
claude plugin install g1t@g1t
sh

Then work as you always do in a g1t repository: the session shows on the project's page as it runs. Getting started walks through it in full, and Connect a coding agent sets up other harnesses. An assistant can do it for you from https://g1t.sh/llms.txt.

What's in it

Coordinating agents

  • Sessions. Every run of a coding agent is recorded as it works: its prompts, tool calls, subagents, branches, commits and pull requests. Work that is not an issue yet is a session titled by what it is for. The Claude Code plugin does it with no setup; any harness can with the CLI or MCP.
  • Live now. A project's page shows who is working on it right now and with which harness, grouped by issue, pull request or branch.
  • Claims. Advisory, expiring claims on issues, so two agents don't do the same work, with hand-offs between sessions and coordination events (g1t work events --follow, or over MCP) when work overlaps.
  • Integration. The merge queue tests changes together before they land, tests smaller combinations to narrow a failure without assuming a single culprit, and sends the evidence to the session responsible for repair.
  • Why-blame and decisions. g1t why file:line answers why code is as it is: its commit, pull request, issue, the decisions recorded along the way, and the sessions behind it.
  • Ask g1t. Ask about your work from any page, or by messaging @g1t. It answers with your workspace's own AI provider and reads g1t as you.

Code

  • Public and private repositories, git over HTTPS, creating a repository by pushing to it, GitHub import (every branch and tag, and issues).
  • Issues, milestones, labels, pull requests, line comments and reviews.
  • Required checks, rulesets, CODEOWNERS and the merge queue.
  • Workflows from .g1t/workflows in GitHub Actions syntax, on g1t's Linux runners or self-hosted ones on Linux, macOS and Windows, with caches, secrets and variables.
  • Packages: npm, Cargo, Composer, Maven, NuGet, RubyGems, Go modules and container images.
  • Secret push protection, history scanning and dependency update pull requests.

Deployments and observability

  • A preview of every pull request on g1t.page, production on merge, custom domains. Static sites, Workers and containers, chosen from the project.
  • Logs, traces, metrics and errors from your apps over OpenTelemetry.

The workspace

  • Chat: channels, direct messages and threads, live.
  • Artifacts: docs written together, live, with comments, suggestions and history.
  • Home, notifications with browser push, people and teams, roles, invites.
  • A desktop app for Windows and Linux that can also be a runner for your workspace, and an Android app in testing (g1t.sh/download).

Extend

  • REST API, OpenAPI, an MCP server, webhooks and the g1t CLI, all from one set of operations.
  • Integrations: your own AI providers, Sentry, Datadog and webhook alerts, Jira and Linear.
  • Usage billing with no seats and a public price book.

Layout

PathWhat it isLanguage
apps/webThe site: server-rendered React on a Worker. Holds no data.TypeScript
apps/docsThe documentation site, with the API explorer.TypeScript
apps/apiREST API and MCP server.Rust
apps/desktopThe desktop app (Tauri), with a runner inside.Rust, TypeScript
apps/mobileThe phone app (Expo).TypeScript
apps/statusThe status page at status.g1t.sh.TypeScript
apps/sudog1t's own staff console.TypeScript
services/identityAccounts, workspaces, sessions, keys, tokens and invites.Rust
services/reposRepository registry, contents, forks, diffs, landing, git over HTTPS.Rust
services/workIssues, pull requests, reviews, check runs, agent sessions, claims and the merge queue.Rust
services/eventsThe event bus and its log.Rust
services/searchSite-wide search and Explore.Rust
services/billingUsage, the price book, limits, invoices and payments.Rust
services/actionsWorkflows, runs, caches and self-hosted runners.Rust
services/securityPush protection findings, history scanning and dependency updates.Rust
services/integrationsAI providers, Ask, alerts, trackers and the GitHub App.Rust
services/webhooksWebhook deliveries.Rust
services/packagesThe package registries and container images.Rust
services/runnerStarts sandboxes for workflow jobs, builds and the merge queue.TypeScript
services/projectsProjects: what a workspace builds and runs, and where its code lives.TypeScript
services/deploymentsBuilds, previews and production on g1t.page.TypeScript
services/pagesServes every app deployed on g1t.page, and custom domains.TypeScript
services/telemetryObservability: OpenTelemetry ingest, logs, traces, metrics and errors.TypeScript
services/chatChannels, direct messages, threads and their live sockets.TypeScript
services/artifactsDocs: their spaces, sharing and live editing.TypeScript
services/notifyNotifications: each person's feed, live counts and browser push.TypeScript
services/ogSocial cards at og.g1t.sh: a PNG per page, showing only what anyone may see.TypeScript
crates/g1tThe g1t command line.Rust
crates/runnerThe program inside a sandbox, and the self-hosted runner: runs a workflow job or a build and reports back.Rust
crates/contractsTypes and service interfaces for the Rust services.Rust
crates/kitPlumbing shared by Rust services on Workers.Rust
crates/rulesThe rules engine for rulesets on branches and tags.Rust
crates/blobstoreObject storage: R2 on Cloudflare, any S3-compatible store self-hosted.Rust
crates/actionsReads workflows and evaluates their expressions.Rust
crates/scanSecret and lockfile scanning, shared by services.Rust
crates/secretsSecrets at rest and signatures.Rust
crates/sshdGit over SSH, bridged to Cloudflare Artifacts. Not deployed.Rust
packages/contractsThe same interfaces for TypeScript callers.TypeScript
packages/themeDesign tokens and the logo, shared by the site and the docs.CSS
plugins/claude-codeg1t's Claude Code plugin: hooks, skills and commands, published to flagon-io/g1t-plugins on every change.JavaScript
plugins/other-harnessesThe AGENTS.md snippet for Codex, Cursor and other harnesses.Markdown
deploystack.jsonc, every deployable part and its resources; self-host, the Docker Compose version.JSON, Docker Compose

Each service is its own Worker, and each one that keeps data has its own database. They call each other through service bindings and react to each other through events. The core services are written in Rust; the web apps and the rest of the Workers in TypeScript.

Run your own

On your own machine

The core forge runs in Docker, with no Cloudflare account:

docker compose -f deploy/self-host/docker-compose.yml up --build
sh

Then open http://localhost:8787 and sign up. The confirmation mail is in Mailpit at http://localhost:8025. Repositories, push and clone, issues, pull requests, code browsing, search, webhooks and observability work, and the API and MCP server answer at http://localhost:8789; packages and container images are kept in the bundled S3-compatible store, RustFS. Deployments and billing are off, and the merge queue can't test or land pull requests without g1t's runner. Run g1t yourself says what works, and how a self-hosted g1t runs what runs inside it.

On Cloudflare

You need a Cloudflare account on the Workers Paid plan (Cloudflare Artifacts, where repositories are stored, requires it), Node 24 (.nvmrc, and engines in package.json), Rust with the wasm32-unknown-unknown target, and Docker to build the sandbox image.

npm install
npx wrangler login
sh

Then, once:

  1. Create the resources each part needs: D1 databases, queues, KV namespaces, R2 buckets and the Cloudflare Artifacts namespace (npx wrangler d1 create <name>, npx wrangler queues create <name>, and so on), and set each part's secrets. deploy/stack.jsonc lists them all.
  2. Put your own account_id, database ids and hostnames in each wrangler.jsonc.
  3. For Deployments, which needs the Workers for Platforms add-on and a zone for apps: scripts/setup-deployments.sh.

Deploy everything, migrations first, in dependency order:

scripts/deploy.sh
sh

Or only what changed, still in order: scripts/deploy.sh billing web. Both use your wrangler login, not a token in .env.

Create the first account by registering on your site, or with node services/identity/scripts/create-user.mjs <username>. Deploy g1t to Cloudflare covers the deploy tool, the workflow and first-time setup in full.

License

MIT