A persistent background code review database for AI coding agents that auto-triggers reviews via git hooks, with TUI display and agentic auto-fix closed loop.
roborev is a cross-platform CLI tool created by Wes McKinney (creator of Apache Arrow / Pandas), positioned as a persistent background code review database for AI coding agents. After installing a git post-commit hook via roborev init, every commit automatically triggers a background review, with findings displayed in real-time through a terminal TUI.
Core Capabilities#
Automatic Review & Triggering
- Git post-commit hook integration automatically queues review tasks on every commit without workflow changes
- Manual triggering by specific SHA (
roborev review <sha>), full branch (--branch), or uncommitted changes (--dirty) roborev compactvalidates open review findings, filters false positives, and merges related issues
Agentic Fix Closed Loop
roborev fixsends review findings to an AI agent, auto-applies fixes and commitsroborev refineiterates "fix → re-review" until all reviews pass or max iterations reached (--max-iterations, default TBD)- Fixes execute in isolated git worktrees, keeping the working directory clean
Built-in Code Analysis
7 analysis types, all with optional auto-fix:
| Type | Description |
|---|---|
duplication | Duplicate code detection |
complexity | Complexity analysis |
refactor | Refactoring suggestions |
test-fixtures | Test fixture issues |
dead-code | Dead code detection |
api-design | API design review |
architecture | Architecture-level review |
Multi-Agent Compatibility
Auto-detects and supports 10+ AI coding agents: Codex, Claude Code, Gemini CLI, Copilot, OpenCode, Cursor, Kiro, Kilo, Droid, Pi, etc. Supports routing agents like Claude Code to local model endpoints (Ollama, LiteLLM), with separate model specification for review and fix.
Interactive Terminal UI
Real-time review queue, vim-style keyboard navigation, Markdown rich-text rendering (syntax-highlighted code blocks, headings, lists), configurable color themes (auto / dark / light / none).
Architecture Highlights#
- Daemon: Local HTTP server, default port 7373, auto-finds available port if occupied
- Worker Pool: 4 parallel review workers (configurable)
- Local Storage: SQLite (WAL mode), path
~/.roborev/reviews.db - Multi-Machine Sync: Bidirectional sync via PostgreSQL, local SQLite as fast cache
- Configuration: Global
~/.roborev/config.toml+ repo-level.roborev.toml - Code Structure:
cmd/roborev(CLI entry),internal(core logic),skills(agent skills),packaging/systemd(service packaging)
Review Trigger Flow:
git commit → post-commit hook → daemon HTTP API → worker pool queues → results written to SQLite
Fix Execution Flow:
roborev fix/refine → create isolated git worktree → invoke local AI agent CLI → agent modifies code in worktree → review validation → merge back to main branch
Installation & Quick Start#
# Shell Script (macOS / Linux)
curl -fsSL https://roborev.io/install.sh | bash
# Homebrew (macOS / Linux)
brew install roborev-dev/tap/roborev
# Windows (PowerShell)
powershell -ExecutionPolicy ByPass -c "irm https://roborev.io/install.ps1 | iex"
# Go install
go install github.com/roborev-dev/roborev/cmd/roborev@latest
cd your-repo
roborev init # Install post-commit hook
git commit -m "..." # Review auto-triggers
roborev tui # Open interactive review UI
CLI Commands#
| Command | Description |
|---|---|
roborev init | Initialize in current repo |
roborev tui | Open interactive terminal UI |
roborev status | Show daemon and queue status |
roborev review <sha> | Queue specific commit for review |
roborev review --branch | Review all commits on current branch |
roborev review --dirty | Review uncommitted changes |
roborev fix | Fix open reviews (optional job ID) |
roborev refine | Auto-fix loop |
roborev analyze <type> | Run specific code analysis type |
roborev compact | Validate and merge review findings |
roborev show [sha] | Show review results for a commit |
roborev run "<task>" | Execute custom task with AI agent |
roborev close <id> | Close a specific review |
roborev skills install | Install Claude/Codex agent skills |
Additional Features#
- Extensible Hooks: Run custom shell commands on review events; built-in beads integration auto-creates trackable issues from review failures
- Fully Local: No hosted service or extra infrastructure needed; all review orchestration runs locally
- Key Environment Variables:
ROBOREV_DATA_DIR(override default data dir),ROBOREV_COLOR_MODE(TUI color theme),NO_COLOR(disable colors),ROBOREV_CLAUDE_PROXY_TOKEN(forward bearer token to agent endpoint)
Written in Go (98.9%), MIT licensed, currently at v0.54.1 (94 releases, 571 commits).