About Vigil

AI agents write code faster than any team can review. Claude Code, Cursor, Devin — they ship PRs in minutes, complete with confident descriptions. But who verifies that the description matches the diff? Vigil does. We read every claim, check it against the actual code changes, and surface what nobody mentioned. The verification layer between AI-generated code and your main branch.

What we do

Vigil gives every pull request a verification report — claims checked against the diff, undocumented changes surfaced, and structural impact analyzed. Three layers of verification: Trust Verification confirms the PR does what it says and catches what the description missed. Deep Analysis detects contract violations and performs granular diff gap detection. Developer Assist evaluates risk patterns and generates descriptions when the author didn’t write one. Results appear directly on the PR.

How we're different

We complement code review tools, not replace them. CodeRabbit reviews code quality. GitHub Actions runs CI. Codecov measures coverage. Vigil verifies that what your PR claims matches what the code actually does. When your PR says "adds rate limiting" — we check the diff. When it doesn't mention a new Redis dependency — we flag it. Different job, same pull request.

Open source

Vigil is open source under the MIT License. The entire codebase — 840+ tests, 8 signals, the score engine — is public on GitHub. You can self-host, audit, contribute, or fork it.