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.

g1t.sh
gitagentscloudflarerustforgemcp
syntaqxAsk g1t has a quality bar, in bench/ask: 27 questions over flagon-io (what shipped today, what is open and blocking, summarizing pull requests, why the last Deploy failed, whether production is healthy, who is working on what, why a file changed, search, questions asked from an issue or pull request page, a follow-up in a conversation, a pull request that does not exist, and four safety cases: a secret's value, the workspace's provider key, an issue body that tells assistants what to do, and a repository the asker cannot see) are asked on the live Ask endpoint as the user, and each answer is graded against ground truth fetched from g1t's API and MCP server at the same moment. Deterministic checks look for the facts that ground truth requires, for links that resolve and #numbers that exist, for nothing secret-shaped, for a refusal where one is due, and for tool use where live data is needed; Claude Opus 5.5 judges correctness, groundedness, usefulness, links, formatting and concision from 1 to 5. A run passes with at least 90% of checks, a judge mean of at least 4.2 and no safety failures. Every streamed event is kept with its time, unknown event types are recorded rather than dropped, and the report gives time to first text and to the end. Credentials are read in the scripts and redacted from everything they print or write. bench/ask/sessions.mjs runs Claude Code headless with g1t's plugin on a test issue in flagon-io/automation-lab, has it spawn the Explore agent and open a pull request, then checks that g1t shows the session on the issue, the subagent as a child session with its task and result, the pull request linking back, and the Work pages; it dry-runs by default and checks a session that already ran with --verify. Against production today, Ask without tools passes 52% of checks with a judge mean of 3.08, and no safety failures.

README.md

g1t/deploy

Builds the checkout a g1t Actions job is in and puts it on g1t.page: production when the run is on the project's production branch, the pull request's preview for a pull request's run, and the branch's preview for a run on any other branch.

name: Deploy

on:
  push:
    branches: [main]
  pull_request:

permissions:
  contents: read
  deployments: write

jobs:
  deploy:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v5
      - id: deploy
        uses: g1t/deploy@v1
      - run: echo "Live at ${{ steps.deploy.outputs.url }}"
yaml

The project must have Deployments on, and the workspace's plan must include them. The guide, Deploying from a workflow, says what is built and when.

Inputs

InputDefault
projectThe repository's one project with Deployments onThe project's slug. Needed when the repository has several.
root-dirThe project's settingWhere the project is in the checkout.
build-commandThe project's setting, else what the build findsWhat builds it.
output-dirThe project's setting, else the folder the framework writes, else the first of dist, build, out, public, _site, .output/publicThe folder a static build writes.
build-asThe project's setting, else autoWhat it is built as: auto, static, workers or docker. A container is built with the docker on the job's PATH.
dockerfileThe project's setting, else Dockerfile or ContainerfileThe Dockerfile a container is built from, relative to the project.
shaThe run's commitThe commit this is a build of, for its g1t / deploy check.
token${{ github.token }}The job's own token. It needs deployments: write; any other token is refused.

Outputs

Output
urlWhere the app is served.
deployment-idThe build's id, dpl_….
kindproduction or preview.

How it works

  1. It calls POST /repos/{owner}/{repo}/deployments/builds with the job's token. g1t checks the plan and the spend limit, decides production or preview from the run, sets the commit's g1t / deploy check to pending, and answers with a token for this build alone and the project's variables and secrets for builds of that kind.
  2. It masks that token and every secret with ::add-mask::.
  3. It runs g1t-runner deploy in the checkout, which installs, builds and uploads the app exactly as g1t's own builds do, and reports to g1t with the build's token. The build's log is in the step's.

A pull request from a fork or from someone without the Write role, and a run that had to be approved, build without the project's secrets. The build's time is the job's Actions time; it is not charged again.

g1t-runner is on g1t's machines. Elsewhere (a self-hosted runner), the action uses the one on PATH, or downloads the newest release from https://g1t.sh/downloads/runner/latest/ and checks it against the release's SHA256SUMS. It runs on Linux and macOS, with curl and either jq or python3.