Hi Product Hunt, I'm Sarath, maker of Co-Review.

I build with a coding agent most days. Each diff looked fine. Six months in, the codebase worked, but I couldn't tell you where things lived or why they were built that way. It had stopped being mine.

So I built the tool I wanted: open the entire repository as a review, the way you'd read a pull request, with the agent that wrote it sitting in the thread. Comment on a line, a function, a file or a whole folder and ask why. The agent answers from the code, not from memory of a chat.

What it does:

Whole-repository audit: the agent writes a guide to the map, vocabulary, flows and weak spots, with findings grouped by area. Coverage tracks what you've actually read.
Your change as a PR page: line comments, suggested edits, one verdict, and the agent acts on it.
Someone else's PR: the agent takes the first pass, you decide what goes back to the author, and it posts to GitHub as a real review with line comments.
Design before code: an L1/L2/L3 plan with Mermaid diagrams, revised from your comments before anything is built.

It's a review tool with IDE features, built on Eclipse Theia, so navigation, search, Git and language servers are the real thing and everything unrelated to reviewing is gone.

Made for Claude Code (one plugin install). Native packages for Pi and Oh My Pi. Any MCP agent works: Codex, Cursor, Gemini CLI, Zed, OpenCode, Goose.

Free and MIT licensed. No account, no telemetry. Reviews live on your machine; your code goes only to the agent you chose.

Two-minute walkthrough:

Install:
curl -fsSL https://sdsvn.github.io/co-revie... | bash

I'd love to hear what you'd want to ask your own codebase. I'll be here all day.