CI & GitHub

Two ways to gate pull requests on model lifecycle: run mm ci in any CI system, or install the GitHub App for a one-click setup.

mm ci in any pipeline

# .github/workflows/mm.yml
- run: npx @modelstatus/cli ci . --fail-on retiring

mm ci scans the directory against the signed registry and exits non-zero when any usage is at or above the --fail-on threshold:

It runs offline after the first registry fetch and needs no account. Under GitHub Actions it emits inline annotations on the offending lines and a job step summary. Useful flags:

• --diff <base> — only report findings in files changed vs a base ref. On pull requests this is automatic via GITHUB_BASE_REF.
• --json-out <file> — write clean findings JSON to a file (stdout stays annotations-only).
• --report — (Pro) sync the run’s usages and a CI-run row to your cloud account, so the dashboard tracks drift per branch.

Not on GitHub? The same command works in GitLab CI, CircleCI, Jenkins, or a Dockerfile — the exit code is the contract. Binaries are on the CDN (manifest: cli/latest/version.json) if you’d rather not pull from npm.

The GitHub App

Install the LLM Status GitHub App  and every pull request gets a check run — no workflow file, no secrets, no CLI in your pipeline.

• The check fails with a findings table when a PR introduces a deprecated or retiring model.
• When a finding has a known replacement, the check-run page shows an Open fix PR button. Click it and the app opens a pull request against your branch with the exact rewrites, shown as a diff.
• The app reads only what it needs to scan and posts results back as checks and comments.

The App requires a Pro account.

Free vs Pro

Details on tiers: Free vs Pro.