kenn-io/roborev

roborev – Continuous AI‑powered code review for coding agents

What it is

• A Go‑based command‑line tool that installs a post‑commit Git hook and a background daemon. Every time a commit is made, roborev sends the changed code to an LLM‑based coding agent (Claude Code, Codex, Gemini, Copilot, etc.) and receives a review.
• The review is shown locally in a terminal UI (TUI) or a native browser UI. It can also be fed back to the same agent so the agent can automatically apply fixes.

Why it matters

• Developers who rely on AI coding assistants often get suggestions that later turn out to be buggy. roborev closes the loop by reviewing each commit immediately, before the changes accumulate.
• Because everything runs on the developer’s machine, no external service or cloud infrastructure is required, keeping latency low and data private.

Key features (as described in the README)

Feature	What you get
Background reviews	A Git hook triggers a review for every commit, without needing a remote CI service.
Auto‑fix / refine loop	roborev fix sends the review back to the agent, which applies a commit; roborev refine repeats this until the review passes.
Agent hook	Optional integration with nine popular AI coding agents (Claude Code, Codex, Gemini, Copilot CLI, Cursor, Factory Droid, Hermes, Qwen, Pi). The hook can surface open findings directly inside the agent’s UI.
Built‑in analyses	Duplication, complexity, refactoring, dead code, security, test‑fixture suggestions, API‑design, architecture checks, etc. Each analysis appears as a review that can be auto‑fixed.
Interactive TUI	Vim‑style navigation, split‑screen view of the review queue and details.
Native browser UI	A local web UI that shows the queue, lets you filter, comment, close, or rerun reviews, and displays analytics (cost, latency, reliability).
Review verification	roborev compact de‑duplicates findings and filters false positives before they are presented.
Extensible hooks	Run arbitrary shell commands on review events; built‑in beads and kata integrations can automatically create issue tickets.
Export & analytics	Export completed reviews, CI‑panel metrics, and cost data as JSON for reporting or archival purposes.
Configuration	.roborev.toml (repo‑local) or ~/.roborev/config.toml (global) to set default agent, model, review guidelines, commit author info, and kata integration settings.

Typical workflow

1. roborev init – installs the post‑commit hook.
2. Make a commit as usual; the hook queues a review.
3. Run roborev tui (or roborev ui) to see the findings.
4. Either fix manually, or run roborev fix / roborev refine to let the attached AI agent apply the suggested changes automatically.
5. Optionally install roborev agent‑hook so the agent’s own UI can surface open findings in real time.

Installation options

• One‑line shell script (curl … | bash)
• Homebrew tap (brew install kenn-io/tap/roborev)
• PowerShell script for Windows
• go install go.kenn.io/roborev/cmd/roborev@latest
• Building from source with make install for the embedded browser assets.

Who it’s for

• Teams that already use AI coding assistants and want an automated, local code‑review step.
• Solo developers who want immediate feedback on AI‑generated code without configuring a CI pipeline.
• Anyone needing a reproducible audit trail of AI‑generated changes (exports, cost tracking, issue‑creation hooks).

Bottom line roborev is a genuine open‑source project that stitches together Git hooks, local LLM agents, and interactive UIs to provide continuous, AI‑driven code review and automated fixing—all running on the developer’s machine. It fits squarely in the frontier of AI‑augmented software development tools.