Term MCP Deepseek

🚀 DeepSeek MCP Type Terminal Server

This project implements a prototype of an MCP protocol-like server using the DeepSeek API. It aims to demonstrate the core concepts of MCP by exposing endpoints that allow an AI assistant to perform the following operations:

• List available tools
• Invoke commands on an active Shell session
• Integrate an AI chat (DeepSeek) that can include special instructions (e.g., CMD: lines) to trigger command execution

⚠️ Important Note

Although this implementation captures many core concepts of MCP, it is not yet a fully compliant MCP server as defined by Anthropic. It is a proof-of-concept design and requires further improvements (e.g., JSON-RPC protocol support, real-time updates via SSE, session management, and improved security) before it can be used in a production environment.

✨ Features

• Chat Interface: A simple web-based chat client using Flask and Tailwind CSS, where users can interact with the server.

• AI Integration: Use the DeepSeek API to generate responses. The AI can instruct the server to execute terminal commands by including lines starting with CMD:.

• Terminal Command Execution: Execute Shell commands by maintaining a persistent Bash session using the pexpect library and returning the output to the client.

• MCP Endpoints: Provide /mcp/list_tools and /mcp/call_tool endpoints, mimicking the MCP tool discovery and invocation functions.

🚀 Quick Start

Prerequisites

• Python 3.8+
• pip
• A valid DeepSeek API key

📦 Installation

1. Clone the repository:

   git clone https://github.com/OthmaneBlial/term_mcp_deepseek.git
   cd term_mcp_deepseek
   

2. Create and activate a virtual environment:

   python3 -m venv venv
   source venv/bin/activate  # On Windows, use `venv\Scripts\activate`
   

3. Install the required dependencies:

   pip install -r requirements.txt
   

4. Configure your API key:

   Update DEEPSEEK_API_KEY in .env with your DeepSeek API key.

Running the Server

Run the Flask server:

python server.py

Access http://127.0.0.1:5000 to access the chat interface.

📚 Documentation

Chat Endpoint

• URL: /chat
• Method: POST
• Payload: { "message": "Your message content" }
• Description: Add the user's message to the conversation, send it to the DeepSeek API, look for any command instructions (CMD:), execute them, and return the final response.

MCP Endpoints

List Tools

• URL: /mcp/list_tools
• Method: POST
• Response: A JSON list showing available tools (e.g., write_to_terminal, read_terminal_output, send_control_character).

Call Tool

• URL: /mcp/call_tool
• Method: POST
• Payload:

  {
    "name": "tool_name",
    "arguments": { ... }
  }
  

• Description: Invoke the specified tool and pass parameters as needed.

🔧 Future Improvement Directions

1. Protocol Standardization: Implement full support for the JSON-RPC protocol to ensure compatibility with existing systems.

2. Real-time Communication: Integrate WebSocket or Server-Sent Events (SSE) to enable real-time command execution and output feedback.

3. Session Management: Add user authentication and session management features to support multiple simultaneously connected clients.

4. Error Handling and Logging: Implement more robust error handling mechanisms and integrate logging to track server operations and debug issues.

5. Performance Optimization: Optimize the server code and configuration to improve command execution speed and overall response time.