DeepSeek MCP Server

Example Use Cases

• 1

  Connect MCP-compatible applications to DeepSeek models while preserving multi-turn context and conversation state.

• 2

  Configurable generation controls (temperature, max tokens, top_p, presence/frequency penalties) and runtime model selection.

• 3

  Resilient model serving with automatic fallback (deepseek-reasoner -> deepseek-chat) and local testing via the MCP Inspector.

Description

This project provides a Model Context Protocol (MCP) server that exposes DeepSeek models to MCP clients, preserving conversation history and configuration across multi-turn dialogues. It supports model discovery, runtime configuration (temperature, max tokens, top_p, penalties), and automatic fallback from the primary reasoner model to a general-purpose chat model. The server can be run locally or integrated into clients like Claude Desktop and tested with the MCP Inspector, making it useful for both interactive use and data collection for model training.