Home / MCP / DeepSeek MCP Server

DeepSeek MCP Server

Provides an MCP interface to DeepSeek API for Claude Desktop with configurable models, temperature, token limits, and multi-turn support.

javascript

2stars

View on GitHub

Installation

Add the following to your MCP client configuration file.

Configuration

View docs

{
    "mcpServers": {
        "deepseek": {
            "command": "npx",
            "args": [
                "-y",
                "deepseek-mcp-server"
            ],
            "env": {
                "DEEPSEEK_API_KEY": "YOUR_API_KEY"
            }
        }
    }
}

DeepSeek MCP Server provides an MCP-compatible proxy that lets Claude Desktop connect to DeepSeek models through a single, configurable interface. You can control model selection, temperature, tokens, and multi-turn conversation behavior while keeping your API key secure and abstracted behind the MCP layer.

How to use

To use the DeepSeek MCP Server with an MCP client such as Claude Desktop, run the server locally and point Claude Desktop to the local MCP endpoint. You authenticate with your DeepSeek API key, which is provided to the server via environment variables.

Configure Claude Desktop to load the MCP server as a stdio-driven process. The server runs as a local helper that you start from your command line, and Claude Desktop communicates with it through the standard input/output transport.

How to install

Prerequisites: you need Node.js and npm installed on your machine.

Option A: Install manually via npm

npm install -g deepseek-mcp-server

Option B: Install via Smithery for automatic setup with Claude Desktop

npx -y @smithery/cli install @dmontgomery40/deepseek-mcp-server --client claude

After installation, you will configure Claude Desktop to launch the MCP server as a stdio process using the command and environment shown in the configuration example.

Configuration notes

Set up Claude Desktop to start the MCP server as a local stdio process with the following settings. Replace your-api-key with your actual DeepSeek API key.

{
  "mcpServers": {
    "deepseek": {
      "command": "npx",
      "args": [
        "-y",
        "deepseek-mcp-server"
      ],
      "env": {
        "DEEPSEEK_API_KEY": "your-api-key"
      }
    }
  }
}

Testing and debugging

To verify the setup, build and run the MCP Inspector to connect to your local server. This helps you view available tools, test chat completions with configurable parameters, and monitor performance.

npm run build

# Then test with MCP Inspector
npx @modelcontextprotocol/inspector node ./build/index.js

Available tools

models

Query the available DeepSeek models and their capabilities, enabling you to select the appropriate model for a given task.

model-config

Fetch and adjust configuration options such as temperature, max tokens, top_p, presence penalty, and frequency penalty for model responses.