概览
What is AI Customer Support Bot - MCP Server?
AI Customer Support Bot - MCP Server is a Model Context Protocol (MCP) server that provides AI-powered customer support by integrating with Cursor AI and Glama.ai. It is designed for developers and support teams who need real-time context fetching, batch processing, and intelligent response generation within an MCP-compliant framework.
How to use AI Customer Support Bot - MCP Server?
Install Python 3.8+, PostgreSQL, and obtain API keys for Glama.ai and Cursor AI. Clone the repository, set up a virtual environment, install dependencies, copy .env.example to .env and fill in credentials, create the database, run migrations, then start the server with python app.py. The server runs on http://localhost:8000 and exposes endpoints like POST /mcp/process for single queries and POST /mcp/batch for batch processing.
Key features of AI Customer Support Bot - MCP Server
- Real-time context fetching from Glama.ai
- AI-powered response generation with Cursor AI
- Batch processing and priority queuing
- Rate limiting (100 requests per 60 seconds)
- Health monitoring and error tracking
- MCP protocol compliance
Use cases of AI Customer Support Bot - MCP Server
- Automatically answering customer inquiries (e.g., password reset instructions)
- Processing multiple support tickets in a single batch request
- Monitoring server health and rate limit usage in production
- Integrating Glama.ai and Cursor AI into an existing MCP workflow
- Handling high-priority requests with priority queuing
FAQ from AI Customer Support Bot - MCP Server
What are the prerequisites to run the server?
You need Python 3.8+, a PostgreSQL database, and valid API keys for Glama.ai and Cursor AI.
How do I configure environment variables?
Copy .env.example to .env and set your Glama API key, Cursor API key, database URL, secret key, and server parameters like MAX_CONTEXT_RESULTS and RATE_LIMIT_REQUESTS.
Which API endpoints are available?
The server provides GET / (root info), GET /mcp/version, GET /mcp/capabilities, POST /mcp/process, POST /mcp/batch, and GET /mcp/health. All MCP endpoints require an X-MCP-Auth header.
How does rate limiting work?
By default, the server allows 100 requests per 60 seconds. Rate limit usage is exposed via the health endpoint, and exceeded requests return a RATE_LIMIT_EXCEEDED error with reset time.
What error codes does the server return?
Common error codes include RATE_LIMIT_EXCEEDED, UNSUPPORTED_MCP_VERSION, PROCESSING_ERROR, CONTEXT_FETCH_ERROR, and BATCH_PROCESSING_ERROR. All errors follow a structured JSON format.
