DeepSeek MCP Server

This is a Model Context Protocol (MCP) server designed for seamless integration of DeepSeek's language models with MCP-compatible applications like Claude Desktop, enabling anonymous DeepSeek API usage through a proxy.

Installation

Installation can be done automatically via Smithery or manually using npm. For Claude Desktop, configuration involves adding the server details to claude_desktop_config.json, including the DeepSeek API key.

Features

The server supports natural language requests for configuration changes and model queries. It offers automatic model fallback from deepseek-reasoner (R1) to deepseek-chat (v3) if the primary model is unavailable, with the ability to switch between them via prompts. Resource discovery enables custom model selection, temperature control, token limits, and penalty adjustments.

Enhanced Conversation Features

Multi-turn conversation support maintains context across exchanges, which is valuable for training/fine-tuning DeepSeek models and managing complex interactions in production, such as multi-step reasoning, troubleshooting, and technical discussions.

Testing with MCP Inspector

The server can be tested locally using the MCP Inspector tool by building the server and running it with the inspector, allowing you to view tools, test chat completions, debug responses, and monitor performance. The server defaults to DeepSeek's R1 model (deepseek-reasoner).

License

MIT