# Review Assist MCP

> Review Assist MCP is an open-source MCP server for developer tools. Turn a coding agent's session into an Intent Document checked against the diff. It runs locally on your machine and is started by your AI client.

- **Kind:** MCP server
- **Pricing:** Open source
- **Platforms:** CLI, self-hosted
- **Website:** https://reviewassist.dev
- **Repository:** https://github.com/uditk2/review-assist/tree/HEAD/packages/mcp-server
- **MCP Registry name:** io.github.uditk2/review-assist-mcp
- **Categories:** developer tools
- **Listing updated:** 2026-11-14

## About

The package targets a review problem: reading AI-written code tells you what changed but not why. Review Assist has the agent write down the real problem, its assumptions, rejected approaches and a walkthrough anchored to each change, then checks that document against the diff before it is saved.

It splits authoring between two roles that cannot reach each other's tools. The Author holds the session transcript but cannot submit; the Intent Reviewer never sees the transcript and is the only one who can submit. The server never calls a model itself; it supplies the diff and transcript and gatekeeps the result.

The server is half of the setup. A separate GitHub App turns the committed document into a guided review on the pull request, with no workflow file and no runner minutes. The site says the session transcript stays on your machine and the app keeps no database, though code passes through the hosted app in transit unless you self-host.

## Features

- **Intent Document:** Captures the problem, assumptions, rejected approaches and anchored walkthrough of changes.
- **Diff validation:** Checks the document against the real diff before writing it.
- **Two-role split:** Author and Reviewer roles are enforced by separate tool sets.
- **Paged diff reading:** Large changes are read a page at a time so responses stay small.
- **GitHub App pairing:** A companion app renders the document as a guided pull request review.
- **No model calls:** The server supplies ground truth and leaves generation to your agent.

## Good for

- Review a large agent-written pull request by its reasoning
- Hand a reviewer the assumptions behind a change
- Audit whether tests cover a stated behaviour
- Keep session context attached to a PR

## Install

### Claude Code (Terminal)

```sh
claude mcp add review-assist-mcp -- npx -y review-assist-mcp
```

### Claude Desktop (claude_desktop_config.json)

```json
{
  "mcpServers": {
    "review-assist-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "review-assist-mcp"
      ]
    }
  }
}
```

### Cursor (~/.cursor/mcp.json)

```json
{
  "mcpServers": {
    "review-assist-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "review-assist-mcp"
      ]
    }
  }
}
```

### VS Code (.vscode/mcp.json)

```json
{
  "servers": {
    "review-assist-mcp": {
      "type": "stdio",
      "command": "npx",
      "args": [
        "-y",
        "review-assist-mcp"
      ]
    }
  }
}
```

### Windsurf (~/.codeium/windsurf/mcp_config.json)

```json
{
  "mcpServers": {
    "review-assist-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "review-assist-mcp"
      ]
    }
  }
}
```

### Other clients (your client's MCP settings (most use this shape))

```json
{
  "mcpServers": {
    "review-assist-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "review-assist-mcp"
      ]
    }
  }
}
```

## Tools

- `compute_diff`: Returns the run handle, numbered hunk index and SHAs for a change.
- `read_diff`: Returns the change itself a page at a time.
- `submit_document`: Submits the validated Intent Document; available to the Reviewer role only.
- `read_transcript`: Gives the Author role the session transcript.
- `get_role_definitions`: Installs the Author and Reviewer roles as subagents.
- `manage_consent`: Manages consent settings.
- `import_session`: Imports a session for authoring.

## FAQ

### Which agents work with it?

It ships tested with Claude Code, the Claude desktop app and Codex, and any MCP client can register it.

### Is it free?

Yes. The site says the server and GitHub App are free under Apache-2.0 with no paid tier.

### Is my code stored?

The site says no: the app reads the document and diff per request and holds them in memory, but code does pass through the hosted service unless you self-host.

### Do I need the GitHub App?

Without it the document is still written to .intent/ in the repo, but nobody sees it on the pull request.

### Is Review Assist MCP free?

Yes. Review Assist MCP is open source, so you can use and self-host it for free.

### How do I add Review Assist MCP to Claude Code?

Run this command in your terminal: claude mcp add review-assist-mcp -- npx -y review-assist-mcp

### What are the best alternatives to Review Assist MCP?

The closest alternatives to Review Assist MCP are REA, Codebase Memory, GitHub, Serena MCP: the IDE for your agent and XERJ.

## Alternatives

- [REA](https://ai-powered.com/mcp/rea): REA is an open-source MCP server for developer tools.
- [Codebase Memory](https://ai-powered.com/mcp/codebase-memory): Codebase Memory is an open-source MCP server for knowledge and memory.
- [GitHub](https://ai-powered.com/mcp/github-2): GitHub is an open-source MCP server for developer tools.
- [Serena MCP: the IDE for your agent](https://ai-powered.com/mcp/serena-mcp-the-ide-for-your-agent): Serena MCP: the IDE for your agent is an open-source MCP server for developer tools.
- [XERJ](https://ai-powered.com/mcp/xerj): XERJ is an open-source MCP server for knowledge and memory.
- [Ouroboros](https://ai-powered.com/mcp/ouroboros): Ouroboros is an open-source MCP server for developer tools.
- [Claude Code Ultimate Guide](https://ai-powered.com/mcp/claude-code-ultimate-guide): Claude Code Ultimate Guide is an open-source MCP server for ai models and llm tooling.
- [Concord MCP](https://ai-powered.com/mcp/concord-mcp): Concord MCP is a freemium MCP server for developer tools.

---
Source: https://ai-powered.com/mcp/review-assist-mcp
