Show HN: Crit – Review AI agent work like you review PRs

Reviewing agent output in a terminal is painful. When Claude generates a 1000-line markdown plan, you can't easily say "change this specific bullet" without copying the text, switching to the chat, and hoping the agent understands the reference. When it updates the plan, you start over from scratch.

Crit is a single-binary CLI that opens your file in a browser with GitHub-style inline comments. You leave your feedback, hit Finish, and a structured prompt goes to your clipboard. Paste it back to your agent.

When the agent edits the plan, Crit shows you a diff between rounds. You can see exactly which points it addressed - no more re-reading the whole thing to figure out what changed.

The same workflow handles code review. Once you're in the implementation phase, crit opens all your changed files with syntax highlighting and line-level comments. Similar interface inspired by GitHub PR review, same clipboard prompt at the end.

To try it out:

---

brew install tomasz-tomczyk/tap/crit

crit plan.md

---

Runs on localhost. No MCP that adds to your context - it comes with optional `/crit` command that helps your agent launch it, but it works perfectly fine without it.

GitHub: https://github.com/tomasz-tomczyk/crit

Homepage: https://crit.live