MCP Deepseek Agent

MCP server implementation using Ollama's Deepseek model for seamless AI integration.

Features

🤖 MCP protocol compliance
🔄 Ollama integration with Deepseek model
⚙️ Automatic configuration
🧹 Clean responses (removes thinking tags)
📝 Standard MCP protocol endpoints

Quick Start

Install Ollama and Deepseek model:

ollama run deepseek-r1

Install the package:

pip install git+https://github.com/freebeiro/mcp-deepseek-agent.git

Start the server:

mcp-deepseek-agent

Configuration

Set through environment variables:

export OLLAMA_API_URL="http://localhost:11434"
export OLLAMA_MODEL="deepseek-r1"
export TEMPERATURE="0.7"
export TOP_P="0.9"
export MCP_PORT="8080"

Usage in MCP Configuration

Add to your MCP configuration:

{
  "mcpServers": {
    "deepseek": {
      "command": "mcp-deepseek-agent",
      "args": [],
      "env": {
        "OLLAMA_MODEL": "deepseek-r1"
      }
    }
  }
}

Documentation

See DOCUMENTATION.md for detailed usage and API documentation.

License

MIT License - see LICENSE file for details.