自作の Model Context Protocol (MCP) クライアントとサーバー

What is 自作の Model Context Protocol (MCP) クライアントとサーバー?

This repository provides a custom implementation of both a Model Context Protocol (MCP) client and server. It connects AI assistants (via Anthropic’s Claude API) with data sources, demonstrating how to build and use MCP for interactive, real‑time communication.

How to use 自作の Model Context Protocol (MCP) クライアントとサーバー?

After cloning the repository, install dependencies with npm install, set up a .env file with your ANTHROPIC_API_KEY, and start the server using npm run start:server. Then run one of the demo applications (npm run demo:http, npm run demo:ws, or npm run demo:chat) to interact with the client.

Key features of 自作の Model Context Protocol (MCP) クライアントとサーバー

• Express.js based RESTful API server
• WebSocket support for real‑time streaming responses
• Unified client supporting both HTTP and WebSocket communication
• Interactive chat console demo with conversation history

Use cases of 自作の Model Context Protocol (MCP) クライアントとサーバー

• Learning and experimenting with the Model Context Protocol
• Prototyping an MCP‑enabled AI assistant that uses Claude
• Building a foundation for custom MCP integrations with streaming

FAQ from 自作の Model Context Protocol (MCP) クライアントとサーバー

What are the runtime requirements?

Node.js version 14 or higher and an Anthropic API key (to access Claude) are required.

What transport protocols does it support?

It supports both HTTP (RESTful API) and WebSocket (for bidirectional streaming).

Where is the conversation data stored?

The demos keep conversation history in memory only. No persistent storage is implemented by default.

Does it include authentication?

No built‑in authentication; the API key is stored in an environment variable (ANTHROPIC_API_KEY). The README advises strengthening security for production use.