git-lrc

git-lrc installs a git hook that runs an AI review on every diff before it becomes a commit. Think of it as a pre-commit code review that runs itself. You don't have to remember to invoke it, and you don't push a separate review step into CI. If a change looks risky, you see it while the context is still fresh.

A wall of inline comments is hard to triage, so git-lrc turns each review into a filterable view. You can sort by Critical, Warning, or Info, drill into categories like Security and Secrets Management, and see live counts of how many issues each area has. That ranking is the point. It tells you what to fix first instead of making you read everything.

The scanner is organized around risk rather than style. Hexmos lists 10 categories and over 100 tracked failure patterns, covering security, reliability, cost, and bugs. It's closer to a code security scanner than a formatting linter. That kind of coverage is more useful to a team worried about outages. Static code review catches what the model misses. That's the whole idea. No dashboard to babysit, either. It just runs.

Every review also produces a short slide deck: what was implemented in plain English, which risks were flagged, and what technical details matter. It's a lightweight way to keep a record of changes without anyone writing a changelog by hand. For onboarding or post-incident reads, that beats re-reading a raw diff.

Findings can be copied, or sent straight to Claude, and fed back into your AI agent to fix. There's also a thumbs up/down on each finding that tunes future reviews, so the signal-to-noise ratio should improve as your team keeps using it.

git-lrc pairs the summary decks with an iteration and coverage history in your git log. The result is a running record of every change and the risks attached to it, which your team can browse months later without maintaining documentation separately.