MCP Server for Cursor Integration

What is MCP Server for Cursor Integration?

A Managed Context Provider (MCP) server that integrates Cursor (the AI coding assistant) with Coda.io documentation. It fetches context from Coda.io tables to provide relevant, context-aware responses within Cursor.

How to use MCP Server for Cursor Integration?

Install Node.js (v14+), clone the repository, run npm install, configure a .env file with your Coda.io API token and document ID. Start the server with npm run dev (development) or npm start (production). Use ngrok (ngrok http 3000) to expose it publicly, then update your MCP client configuration (mcp.json) with the provided ngrok URL and the /context endpoint.

Key features of MCP Server for Cursor Integration

• Connects Cursor to Coda.io documentation via API
• Built-in caching for faster repeated queries
• RESTful API endpoints for context and health
• Supports multiple Coda.io tables simultaneously
• Includes health check and cache management endpoints

Use cases of MCP Server for Cursor Integration

• Query Jira ticket details stored in Coda.io directly from Cursor
• Retrieve API documentation snippets without leaving the editor
• Access architecture decision records while coding
• Combine documentation from several Coda tables into one response

FAQ from MCP Server for Cursor Integration

What does this server do?

It acts as a middleware between Cursor and your Coda.io documentation tables, fetching and returning relevant context based on search queries.

What are the prerequisites?

You need Node.js v14 or higher, a Coda.io account with API access, and a Coda.io document that contains the required tables (Jira Tickets, API Documentation, Architecture Decisions).

How do I expose the server to the internet?

Use ngrok: install it globally (npm install -g ngrok), run ngrok http 3000 while the server is running, and update the url in your mcp.json file with the ngrok‑provided URL.

How do I manage the cache?

Send a POST request to /cache/clear to clear all cache, or a GET request to /cache/stats to view cache statistics.

Which tables does the server expect in Coda.io?

By default it looks for “Jira Tickets”, “API Documentation”, and “Architecture Decisions”. You can change these table names in src/services/mcp.service.js.