Feature

AI Review Loop.

Leave comments, click Finish Review, paste the prompt into your agent. The prompt points the agent to the review file and comments file, and tells it to run crit go <port> when done. Crit detects the signal, reloads with a diff, and you review again.

$ crit plan.md

Listening on http://localhost:3247

→ Round 1: 3 comments added

→ Finish review — prompt copied ✓

→ Waiting for agent…

→ File updated (8 edits detected)

→ Round 2: diff ready — 2 resolved, 1 open

→ 1 new comment added

→ Finish review — prompt copied ✓

→ Waiting for agent…

→ File updated (3 edits detected)

→ Round 3: diff ready — all resolved ✓

How it works

1

When you click "Finish Review", Crit writes a .review.md file (your comments interleaved as blockquotes) and a .comments.json file (structured comment data). A prompt is copied to your clipboard.

2

The prompt tells the agent to address comments in the review file, mark them resolved in the comments JSON, and run crit go <port> when done. Paste it into any agent.

3

When the agent runs crit go, the browser starts a new round with a diff of what changed. Previous comments show as resolved or still open.

4

Add new comments on the updated version and repeat. The loop continues until the plan is right.

Try it out

Install Crit and start reviewing markdown in seconds.

See live demo Star on GitHub

$ brew install tomasz-tomczyk/tap/crit

then run

$ crit plan.md

Or download a pre-built binary from GitHub Releases.

More features

Inline Comments Split & Unified Diff Vim Keybindings Share Reviews Syntax Highlighting Mermaid Diagrams