How it works
A tight loop between your agent and you
Your agent drafts
Claude Code writes a markdown draft — a plan, spec, or summary — and hands it to mdreview instead of charging ahead.
You review in the browser
Read it, leave threaded comments, request changes, or approve. No account needed for local.
The agent revises
It reads your comments back over HTTP or MCP, replies, resolves threads, and pushes a new draft.
Repeat until you approve — the agent never ships past a review you haven't signed off on.
Coding was never the bottleneck. Thinking was.
Agents generate code instantly — but instant, unreviewed output falls over. mdreview puts a plan-and-review checkpoint back in the loop.
Get started
Add mdreview to your agent
Sign in with any email address. There is no invite list. Pick the route that matches your agent.
claude.ai custom connector
No token neededWorks in claude.ai on the web, Desktop and mobile. In claude.ai open Settings, then Connectors, then Add custom connector, and enter this URL.
Claude opens the mdreview sign-in. Enter your email, follow the one-time link, and approve the connection on the consent page. You can revoke it later on your Account page.
Claude Code plugin
Run these two commands inside Claude Code. The plugin asks for a token, which you mint on the Account page after signing in at app.mdreview.space.
Installer
Needs python3 and the claude CLI. It asks whether to run locally or use the hosted instance, and prompts for a token if you choose hosted.
Two ways to run it
Your machine, or ours
Local
Open to everyoneRuns on your machine — no account, no Docker. Starts mdreview on localhost:8137 and wires Claude Code straight to it.
Hosted
Open to everyoneUse the managed instance at app.mdreview.space. Sign in with any email address; you get a one-time link, no password and no invite.