Cursor Review

Executes cursor.review or codex.review prompts and returns structured review JSON for auditing.

typescript

GitHub Stars

typescript

Language

7 months ago

First Indexed

3 months ago

Catalog Refreshed

Documentation & install

Readme and setup notes from the catalogue, plus a client-ready config you can copy for your MCP host.

Installation

Add the following to your MCP client configuration file.

Configuration

View docs

{
  "mcpServers": {
    "kodaimaehata-reviewer-mcp": {
      "command": "./bin/cursor-reviewer-mcp",
      "args": [],
      "env": {
        "CURSOR_API_KEY": "<CURSOR_API_KEY>",
        "REVIEWER_MCP_CODEX_BIN": "/usr/local/bin/codex",
        "REVIEWER_MCP_SCHEMA_DIR": "/path/to/schemas",
        "REVIEWER_MCP_CODEX_FLAGS": "--model \\\"gpt-4\\\" --quiet",
        "REVIEWER_MCP_CODEX_SCHEMA_PATH": "/path/to/codex.schema.json",
        "REVIEWER_MCP_CURSOR_SCHEMA_PATH": "/path/to/cursor.schema.json",
        "REVIEWER_MCP_ALLOW_PLAINTEXT_FALLBACK": "<REVIEWER_MCP_ALLOW_PLAINTEXT_FALLBACK>"
      }
    }
  }
}

You run the MCP server cursor_review to enable automated code review of targets via the Cursor Code tools. This server accepts review requests, executes reviews using Cursor CLI or Codex CLI, and returns only the review JSON. It also saves a timestamped audit of each review for traceability.

How to use

Start the MCP server as a standard stdio service, then send review requests from your MCP client. The server will invoke Cursor Code tools to analyze your targets against a reference and return a structured review JSON. You can configure the client to point to this MCP server and, if needed, supply custom schemas for input validation. All reviews are stored under a timestamped file for auditing.

To perform a review, prepare a request that includes the targets to review and the reference you want to compare against. The server processes the request and replies with a JSON payload representing the review. If you need to reuse the review logic programmatically, you can implement your client to call the MCP endpoint and parse the review JSON from the response.

How to install

Prerequisites: install Node.js 22 LTS and ensure a working shell. You will build and run a local MCP server binary.

Install dependencies and build:

npm i
npm run build

Run the MCP server as a standard input/output (stdio) server:

./bin/cursor-reviewer-mcp

Optionally, register the MCP with Claude Code for project-scoped usage (example):

claude mcp add --scope project cursor-reviewer $(pwd)/bin/cursor-reviewer-mcp

Additional notes

Environment variables are optional in typical local setups. If your execution environment requires authentication for the Cursor client, you may set CURSOR_API_KEY. If you need explicit paths or additional flags for the Codex CLI, you can define REVIEWER_MCP_CODEX_BIN and REVIEWER_MCP_CODEX_FLAGS. You can also override schemas with REVIEWER_MCP_SCHEMA_DIR or specific per-component paths.

The server stores each review in a timestamped JSON file under reviews/. This helps you audit what was reviewed and when.

Available tools

cursor.review

Tool that calls the Cursor CLI to perform a review and returns a review JSON structure.

codex.review

Tool that calls the Codex CLI to perform a review and returns a review JSON structure.

Built by

VeilStrat

AI signals for GTM teams