// the find
HexmosTech/git-lrc
Free, Micro AI Code Reviews That Run on Commit
git-lrc is a commit-time AI code review tool that hooks into git's pre-commit flow, sends your staged diff to an LLM (Gemini by default, BYOK for others), and opens a local browser UI showing inline comments before you finalize the commit. It targets developers using AI agents to write code who want a second pass before anything hits git history. The core idea is sound: catch problems at commit time rather than PR time when the context is still fresh.
The iteration/coverage tracking appended to git log messages is genuinely useful — teams get a transparent record of which commits were reviewed, vouched, or skipped without any separate tooling. The BYOK connector system supporting OpenAI, Claude, DeepSeek, and OpenRouter means you're not locked into Gemini's free tier once you outgrow it. The hook architecture is global-by-default with per-repo disable, which is the right tradeoff — enforcement without constant per-repo setup. The security posture is taken seriously: gitleaks, govulncheck, OSV scanner, Semgrep, and SBOM workflows are all wired into CI, which is more than most tools this size bother with.
The LiveReview cloud service is a required dependency even though the AI key is yours — if that service goes down or changes pricing, the tool breaks regardless of BYOK. The license is 'Sustainable Use' not OSS, so you can self-host but can't build on top of it or offer it internally as a service, which will block adoption in some enterprises. Only the staged diff is analyzed, no repo context — this means the AI can't catch interface mismatches, broken callers, or semantic issues that require understanding what the changed code connects to. The browser UI requirement is awkward in headless environments like containers, CI runners, or remote dev machines where there's no display.