The two review tools split along a real line: one looks at code the agent just wrote, the other at everything that changed. The second is the one that catches problems in the seams — an interaction between two files each of which looked fine alone. Because it runs in the editor, the feedback arrives while the change is still soft, before a reviewer's time is involved. Both tools want repository context passed in, so the call is more than the code itself.
An MCP server that brings Matter AI's code review into the editor, so an AI agent writing code can have that code reviewed in the same session. It runs over a hosted endpoint or locally, and covers review of freshly generated code as well as review of a complete diff.
- A review of code just generated, given the code plus repository owner, name, branch, user and languages — `codereview`
- A comprehensive review over full `git diff` output, with the same repository context — `codereview_full`
- A random cat fact, which is exactly what it sounds like — `cat_fact`
- Tools that auto-discover in Cursor, Windsurf, VS Code and Cline once the server is connected
A Matter AI API key from the settings page of your account. The straightforward path is remote: bridge with `mcp-remote` to the hosted SSE endpoint, passing the key as an `X-AUTH-TOKEN` header. To run it yourself, Python 3.11+ and `pip install -r requirements.txt`, then `python server.py` — the server comes up on `http://localhost:9000` with the SSE endpoint at `/sse`. A Dockerfile is included, exposing the same port. The pyproject name is `matter-ai-mcp` (0.1.7), MIT licensed.
One command plus a key — npx -y mcp-remote https://mcp.matterai.so/sse --header X-AUTH-TOKEN:MATTER_AI_API_KEY, then supply credentials