GitGrit
Define your engineering standards as code — and see how every repository measures up.
GitGrit is a developer-experience platform that gives engineering leaders visibility into how their teams build, and gives developers actionable guidance to improve. Define organization-wide standards as version-controlled Python code, and GitGrit evaluates every repository against them automatically on each push, pull request, and tag — rolling the results into a per-project score and a live badge.
A hosted version runs at app.gitgrit.dev, and the full documentation lives at gitgrit.dev. This repository is the source for the platform, the standard sandbox, the marketplace content, and the public API.
What it does
- Standards as code, not checklists. Express your standards as Python code with full access to repo files, languages, members, branch protection, and history — no DSL to learn.
- Automatic evaluation. Webhooks run matching standards on every push, pull request, and tag, across GitHub and GitLab in one workspace.
- Visible scores. Each project gets a score and an embeddable badge so teams can see where they stand and where to improve.
- Marketplace. Install curated standard packs in one click, then customize.
- Public API. A REST API (OpenAPI at
/api/v1/schema/) for scripts, CI/CD and AI agents — projects, health, commit checks and standards, with every standard running in an isolated sandbox.
The full feature tour is on the docs home.
Quick start
Request access at app.gitgrit.dev; once approved, connect a GitHub or GitLab instance, add a project, then install or write a standard. The full 5-minute walkthrough is in the Quickstart guide.
Writing a standard
A standard is a Python evaluate(project) function that returns a result dict:
def evaluate(project):
if "README.md" not in project.list_files():
return {"passed": False, "score": 0, "message": "No README", "details": {}}
return {"passed": True, "score": 100, "message": "README looks good", "details": {}}
Standards receive a typed ProjectContext, run sandboxed, and ship with mock-data test cases. See the standard guide and the ProjectContext API reference.
Public API
/api/v1/ exposes projects, health per branch, commit checks and standards — the same data as the web app. Authenticate with OAuth (gitgrit auth login in the upcoming CLI), a personal token from your profile page, or a workspace token from the workspace settings (for CI/CD). Interactive docs live at /api/v1/docs/.
Standards drafted or edited by an LLM must be reviewed by a human and run against their test cases before you trust them in production.
Self-hosting
GitGrit is open source under MIT and runs anywhere, including airgapped environments. Requirements: Docker, uv, Python 3.13+.
git clone https://github.com/kfirzvi-com/gitgrit.git
cd gitgrit
docker compose up -d # Postgres
uv sync
uv run poe sandbox # build the standard sandbox image
uv run poe migrate
uv run poe seed # optional — marketplace packs
uv run poe dev # http://localhost:8000
Full instructions and production deployment notes are in the self-hosting docs.
Documentation
Everything lives at gitgrit.dev (source under site/):
Contributing
Contributions are welcome — read CONTRIBUTING.md, sign the CLA, and follow the Code of Conduct. Good starter tasks are labeled good first issue.
Security
Found a vulnerability? Please don't open a public issue — see SECURITY.md for the private reporting process.
License
MIT — see LICENSE.
