MCP Knowledge Base

What is MCP Knowledge Base?

A simple MCP client‑server application that uses OpenAI to answer questions from a built‑in knowledge base. It is intended for learning the MCP protocol or building a custom Q&A system.

How to use MCP Knowledge Base?

Install dependencies with Poetry, set your OpenAI API key in a .env file, start the server via poetry run python server.py, and run the client with poetry run python client-sse.py. The client offers two modes: direct tool calls (uncomment the test line) or LLM‑powered interactions that interpret natural‑language queries.

Key features of MCP Knowledge Base

• MCP server exposing tools via decorators
• MCP client with LLM query interpretation
• Knowledge base stored in data/kb.json
• Direct tool‑call mode for testing
• LLM‑powered mode for natural‑language questions

Use cases of MCP Knowledge Base

• Learn the MCP client‑server architecture interactively
• Build a domain‑specific Q&A system with a custom knowledge base
• Prototype tool calls with or without an LLM orchestrator

FAQ from MCP Knowledge Base

What are the runtime requirements?

Python 3.9 or higher, Poetry for dependency management, and a valid OpenAI API key.

How do I change the knowledge base content?

Edit the data/kb.json file with the Q&A pairs you want to use.

How do I add a new server tool?

Define a new Python function decorated with @mcp.tool() inside server.py.

How do I run the client without an LLM?

Uncomment the asyncio.run(test_direct_tool_calls()) line in client-sse.py to call tools directly.

How do I change the LLM model used?

Modify the model parameter in the MCPClient initialization inside client-sse.py.