Continuous code review for coding agents. Review commits immediately, catch issues early, and fix them while context is fresh.
How roborev works¶ #
roborev reviews every commit in the background and feeds findings back to your coding agent so the write -> review -> fix loop runs hands-off.
Post-commit reviews- every commit is reviewed automatically, with any agent.** Agent hook**- nudges your CLI agent to fix findings mid-session.** Refine before you ship**-/roborev-refine
re-reviews and fixes your whole branch until every review passes, catching bugs before the PR.
Quick Start¶ #
curl -fsSL https://roborev.io/install.sh | bash
Then from within your git repositories:
roborev init # Install post-commit hook
roborev tui # Browse reviews in the terminal UI
For Windows, see the installation guide.
For LLMs¶ #
This docs site publishes source Markdown next to each rendered page. Prefer the
.md
URL when reading or citing docs programmatically: /changelog.md
for
/changelog/
, /guides/reviewing-code.md
for /guides/reviewing-code/
, and
/index.md
for this page.
Why roborev?¶ #
AI coding agents write code fast, but they make mistakes. Most review feedback comes too late. The agent has moved on and context is lost. roborev changes this:
Ask your agents to commit often, ideally every turn of work** roborev reviewseach commit in the background Bring review work back into the agent session**with(agent-hook
--agent droid
for Factory Droid) or check the TUI (roborev tui
) as findings arriveAddress findings by letting the hook prompt the fix skill, copying reviews into your agent, using, or running/roborev-fix
roborev fix
Every commit gets reviewed. Issues surface in seconds, not hours. Open reviews stay in the TUI queue until explicitly addressed and closed, so nothing falls through the cracks.
Review Ledger Every commit is reviewed automatically via git hooks. Reviews accumulate in a persistent queue that acts as a ledger: nothing is closed until explicitly addressed.
Agent-Ready Feedback Use
agent-hook
to prompt Codex, Claude Code, or Droid (--agent droid
) to run the fix skill when review work piles up. You can also copy findings into your agent session, use/roborev-fix
, or runroborev fix
to apply fixes automatically. - Code Analysis Built-in analysis types (duplication, complexity, refactoring, test fixtures, dead code) that agents can address directly.
Multi-Agent Works with Codex, Claude Code, Gemini, Copilot, OpenCode, Cursor, Droid, Kilo, Kiro, Pi, and Grok Build. Auto-detects installed agents.
Rich Markdown Display Reviews render with full Markdown formatting: syntax-highlighted code blocks, headings, lists, and inline styles, right in your terminal.
Runs Locally No hosted service or additional infrastructure. Reviews are orchestrated on your machine using the coding agents you already have configured.
Multi-Machine Sync Bi-directionally sync reviews across machines via PostgreSQL.
Architecture¶ #
Daemon: HTTP server on port 7373 (auto-finds available port if busy)** Workers**: Pool of 4 (configurable) parallel review workers** Storage**: SQLite at~/.roborev/reviews.db
with WAL modeConfig: Global at~/.roborev/config.toml
, per-repo at.roborev.toml
Federated Multiplayer¶ #
Bi-directionally sync reviews across machines with a shared PostgreSQL database. Each daemon maintains its local SQLite for fast access while syncing changes to the central database.