HexmosTech/git-lrc

git‑lrc – AI‑powered micro code reviews that run on every `git commit`

What it is – A command‑line tool (written in Go) that installs a global Git hook. When you stage changes and run git commit, the hook launches a short‑lived UI in your browser that sends the diff to a Large Language Model (Gemini by default, but you can plug in OpenAI, Claude, etc.). The model returns inline comments, risk scores and a one‑page slide deck summarising the change. You can then:

fix the issues and re‑run the review,
“vouch” that you’ve inspected the code and commit anyway, or
skip the review entirely.

All reviews are recorded in the commit message (e.g. LiveReview Pre‑Commit Check: ran (iter:3, coverage:85%)) so the history shows which changes were AI‑checked.

Core features (as described in the README)

Feature	What you get
Automatic pre‑commit review	Hooks run on every commit without extra steps.
Issue Navigator	Structured view of findings across 10 risk categories (security, reliability, cost, etc.) with filters for severity, type and area.
Risk‑scored diff view	Each hunk gets a 0‑100 “blast‑radius” and “review priority” score; the highest‑risk hunks appear first.
Summary Deck	A 60‑second slide deck (plain‑English change description + highlighted risks) generated for each review.
Agent mode	Non‑interactive JSON output for CI or AI‑coding agents (Claude Code, Cursor, …).
Connector manager	UI to add your own LLM API keys – Gemini, OpenAI, Claude, DeepSeek, Anthropic‑compatible, Atlas Cloud, etc.
Feedback loop	Thumbs‑up/down on each finding improves future signal‑to‑noise.
Git‑log tracking	Review status (ran, vouched, skipped) and iteration/coverage stats are appended to the commit message.
Cross‑platform install	One‑liner installer for Linux/macOS (`curl …
Free tier	30 k lines‑of‑code (LOC) per month; premium starts at $32 for 100 k LOC.

How it works (high‑level flow)

Setup – Run git lrc setup (or the one‑liner installer). You provide a LiveReview API key (Hexmos account) and a Gemini API key (or another connector). The tool installs a global Git hook and places a binary under ~/.lrc/bin.
Commit – git add . → git commit … automatically launches the review UI.
Review UI – Shows a GitHub‑style diff with inline AI comments, severity badges, and a risk‑scored list of hunks.
Iterate – Fix the flagged issues, git add . again, and run git lrc review (or simply try to commit again). The tool tracks how many iterations you performed (iter) and what percentage of the final diff has already been AI‑reviewed (coverage).
Finalize – When satisfied, run git lrc review --vouch (records a “vouched” status) or git lrc review --skip (records “skipped”).
Artifacts – A slide deck is saved, and the commit message is annotated with the review status line.

Installation quick‑start

# Linux / macOS (recommended)
curl -L https://hexmos.com/ipm-install | bash && ipm i HexmosTech/git-lrc

# Windows (PowerShell)
iwr https://hexmos.com/ipm-install-ps | iex; ipm i HexmosTech/git-lrc

Alternative direct script: curl -fsSL https://hexmos.com/lrc-install.sh | bash (Linux/macOS) or the PowerShell equivalent.

After install, run git lrc setup to link your LiveReview and Gemini (or other) API keys.

Pricing snapshot

Tier	Free quota	Paid price (starting)	Notes
Free Individual	30 k LOC / month	–	You supply your own AI keys; includes the VS Code extension.
Premium	>30 k LOC	$32 / 100 k LOC (scale by usage)	No per‑seat charge; suitable for small teams.
Enterprise	Custom	Custom	Self‑hosted option, SSO, private domains, tighter data controls.

Who might use it

Engineering teams that generate code with LLMs and want a safety net before the code lands in the repository.
Tech leads / CTOs looking for quantifiable review debt and risk‑based prioritisation across many repos.
CI pipelines that need a non‑interactive review step (git lrc review --agent-mode).
Individual developers who want a free, commit‑time AI reviewer without installing IDE extensions.

Limitations & considerations (from the README)

The tool relies on external LLM APIs; you must provide valid API keys and bear any usage costs.
Scoring (blast‑radius / review priority) needs a local knowledge‑graph binary; it is optional and can be disabled with --blast-radius=false.
The free tier is limited by lines of code processed per month; heavy users will need a paid plan or self‑hosted enterprise deployment.

Bottom line

git‑lrc is a genuine, open‑source (MIT‑style) Go project that brings AI‑driven code‑review directly into the Git commit workflow. It automates risk detection, provides a searchable UI, and records review provenance in the Git history, aiming to “brake” the unchecked changes that LLM‑generated code can introduce.