CodeQL MCP Server

What is CodeQL MCP Server?

It wraps the CodeQL query server and exposes its capabilities via the Model Context Protocol (MCP), enabling tools like Cursor or AI agents to interact with CodeQL through structured commands and doc search.

How to use CodeQL MCP Server?

Install the Python dependencies with uv pip install -r requirements.txt or pip install fastmcp httpx. Start the MCP server using uv run mcp run server.py -t sse, which listens at http://localhost:8000/sse. Configure a client like Cursor by adding the server URL to its .cursor/config.json.

Key features of CodeQL MCP Server

• Register a CodeQL database for querying
• Run full CodeQL queries and retrieve results
• Quick-evaluate a symbol (e.g., a function or class)
• Decode .bqrs result files into JSON
• Locate predicate or class symbol positions in source code

Use cases of CodeQL MCP Server

• Querying a codebase using CodeQL from a Cursor IDE session
• Decoding .bqrs query results into JSON for downstream analysis
• Locating definitions of predicates and classes in code via natural language
• Running CodeQL queries programmatically from an AI agent

FAQ from CodeQL MCP Server

What are the requirements to run the CodeQL MCP Server?

Python packages fastmcp and httpx are required, and a codeql binary must be available in your $PATH (or its path can be hardcoded in codeqlclient.py).

How do I start the CodeQL MCP Server?

Run uv run mcp run server.py -t sse from the project directory. The server will listen on http://localhost:8000/sse by default.

How do I configure Cursor to use the CodeQL MCP Server?

Add a "CodeQL" entry under "mcpServers" in your .cursor/config.json, pointing the "url" to "http://localhost:8000/sse".

What transport does the CodeQL MCP Server use?

The server uses Server-Sent Events (SS