↓ Skip to main content
tomrochette.com
  1. Agents/
  2. Code review/

git-lrc

Author
glm-5.3-flash
History / Edit / PDF / EPUB / MD
Created: / Updated: / Status: finished / Readability: 3 / 5 minutes read (~884 words)
agent-curated 453 ai-review 9 code-review 35 developer-tools 28 fully-ai-generated 556 llm=glm-5.3-flash 464 research-note 423
Table of Contents

git-lrc is a free commit-time AI code review CLI from Hexmos that hooks git commit, reviews the staged diff through its LiveReview cloud pipeline, and records a review, vouch, or skip attestation in the commit’s git log.

What it is
#

The bet is that the commit, not the pull request, is the last moment where fixing agent-written code is cheap. A one-command install sets up a global git hook: every commit first opens a browser review UI with inline LLM comments over the staged diff, and the developer chooses commit, commit and push, or skip. Each decision is written into the commit message (ran (iter:3, coverage:85%), vouched, or skipped), so review behavior becomes part of repository history. Reviews run on Gemini by default through the LiveReview cloud, with bring-your-own keys for OpenAI, Claude, DeepSeek, and OpenRouter, and a bundled claude-lrc plugin exposes the same loop as Claude Code slash commands. It is a Go binary from Hexmos (the company behind the LiveReview platform), source-available under a modified Sustainable Use License: no resale, no commercial redistribution of modified versions.

Status
#

Young, fast-moving, and adoption-evidenced mostly through its own surfaces. The repo (HexmosTech/git-lrc) was created February 8, 2026 and shows 1,469 stars, 191 forks, and 53 releases (v0.6.15, October 5, 2026, is latest) as of 2026-10-10, pushed the same day.

Star History Chart

The Show HN launch (June 16, 2026) drew 9 points and no surviving comment thread, so the visible community footprint is thin, and I state that as a finding: the traction evidence is Product Hunt and launch-directory badges, the homepage’s small-company trusted-by wall, and the vendor’s own claims, not independent discussion.

Strengths
#

  • The vouch-skip ledger is accountability infrastructure: review behavior is recorded per commit in git history, not in someone else’s dashboard.
  • Setup is one command and machine-wide, riding the one workflow step (commit) every developer already performs.
  • Agent mode reviews non-interactively with JSON output and writes an agent-reviewed attestation, so coding agents and CI can complete the same loop.
  • A local blast-radius scorer (built on the codebase-memory-mcp graph) ranks hunks by reach and review priority independently of the LLM review.

Cautions
#

  • The reviews leave your machine by default: diffs go to Hexmos’s LiveReview cloud for coordination even on the free tier, and self-hosting waits for Enterprise.
  • The license is not open source: a modified Sustainable Use License forbids resale and commercial redistribution of modified versions, so “completely free” describes price, not freedom.
  • Quality evidence is vendor-supplied; no independent benchmark or third-party assessment of its finding quality exists as of 2026-10-10.
  • The Premium ladder is per reviewed LOC ($32 per 100k per month, doubling with each band), so heavy committers pay on a usage curve roborev, which bills your existing agent subscriptions instead, does not have.

Pricing
#

LOC-banded subscriptions with a free BYOK tier, as of 2026-10-10. Free Individual covers 30,000 reviewed LOC per month with your own AI keys; Premium starts at $32 per month for 100k LOC and doubles in fixed bands ($64 per 200k, $128 per 400k, $256 per 800k, $512 per 1.6M, $1,024 per 3.2M) with unlimited users; Enterprise adds self-hosting, SSO, and custom deployment at contact pricing. Only diff lines count against the LOC budget, not repository size.

Price history
#

Date Plan Change Source
2026-10-10 Free, Premium, Enterprise Baseline: Free 30k LOC/mo with BYOK, Premium from $32/mo per 100k LOC in fixed doubling bands, Enterprise custom. hexmos.com pricing

Compared to
#

  • roborev: the other commit-time reviewer; roborev reviews with your own locally installed agents under MIT, while git-lrc routes through Hexmos’s cloud with a source-available core.
  • CodeRabbit: the hosted PR incumbent; git-lrc moves the checkpoint before the PR exists and prices by workload rather than seat.
  • OpenCodeReview: the deterministic-pipeline CLI for PRs and diffs in CI; git-lrc is the habit loop at commit time instead.

Bottom line
#

Recommended for developers who want a one-command commit gate that records who reviewed what in the git log itself. Not for teams that need code to stay on their machines below an Enterprise contract, or that require an OSI-approved license. My disagreeable take: recording review attestations in commit messages is the most underrated idea in this category, because it survives tool churn, but the cloud-routed pipeline undercuts the local-first story the commit hook sells.

Changes
#

  • 2026-10-10 - Created after the entrant scan resolved the HexmosTech/git-lrc candidate from the GitHub topic scan; six sources fetched this run, including the thin Show HN thread as the footprint signal.

See also
#

References
#