StackVitals is a free, open source (Apache 2.0) GitHub Action for CI health and DORA metrics.

The problem: when CI goes red on a pull request, someone has to dig through logs to find out if it's a flaky test, something already broken on main, or the change itself. Most teams re-run and hope. And DORA metrics usually mean running a whole server stack.

What it does:

- When CI fails on a pull request, it leaves one comment sorting each failure into likely flaky, already failing on main, or new. A likely flake is re-run, never let through.

- Shows who each open pull request is waiting on: author, reviewer or CI.

- Tracks DORA metrics (deployment frequency, lead time, change failure rate, recovery time), finding where you deploy from with no setup. On 60 public repos it picked the right deployment source for 53.

- Checks dependencies for known vulnerabilities, and tracks how fast newcomers get answers.

No server, no account: the history lives on a branch in your own repo and the dashboard is a static page, so private repos stay private. Run side by side with Apache DevLake on the same 5 repos, DevLake needed 9.2 GB of Docker images and 2h10m; StackVitals took under 10 minutes.

It also reads deployments as CDEvents or from Flux, sends its findings out as events, connects to Jira, Linear, PagerDuty, Slack and GitLab, and includes an MCP server for AI agents.

Setup takes just two commands:
pip install stackvitals
then stackvitals init owner/repo.

Every rule's accuracy is published, misses included.