convez.ai
MCP ServerDevOps & InfraestruturaIntermediário
OthmaneBlial

DeepSeek MCP-like Server Terminal

OthmaneBlial/term_mcp_deepseek

Execute comandos de terminal via chat com DeepSeek através de protocolo MCP-like

Ver no GitHub
InicialPluginScore · de 0 a 100

Servidor tipo MCP que integra a API DeepSeek com execução de comandos de terminal em sessão persistente. Fornece interface web de chat e endpoints REST para descoberta e invocação de ferramentas, com suporte a streaming em tempo real e autenticação básica.

Instalação

$npx convez-ai OthmaneBlial/term_mcp_deepseek

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
18
Forks
9
Acompanhando
1
Saúde
42%
Último commit
há 3 dias
Versão
v1.0.0
Issues abertas
8
Contribuidores
2

Casos de uso

  • Automação de tarefas de terminal via chat com IA
  • Criação de assistentes de DevOps com execução de comandos
  • Prototipagem de servidores MCP para integração com Claude

Veredito

Implementação interessante do conceito MCP focada em execução terminal, mas ainda é um PoC não totalmente compatível com a especificação oficial. Útil para desenvolvedores que querem explorar automação de shell via IA, com atenção necessária à segurança em produção.

README

A tradução para o português deste README ainda está na fila.

DeepSeek MCP-like Server for Terminal

Trust Score This project is an MCP‑like server using the DeepSeek API. It aims to demonstrate the core concepts behind the Model Context Protocol (MCP) by exposing endpoints that allow AI assistants to:

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

Note: While this implementation captures many of the MCP ideas and includes features like real-time streaming, session management, and basic security, it is not yet a fully compliant MCP server as defined by Anthropic. It is designed as a proof-of-concept, and further enhancements (e.g., complete JSON‑RPC protocol support, advanced authentication, and comprehensive error handling) would be needed for production use.

Features

  • Chat Interface: A modern web-based chat client (using Flask and Tailwind CSS) where users can interact with the server, now with real-time updates and improved error handling.

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

  • Terminal Command Execution: Executes shell commands via a persistent Bash session using the pexpect library and returns output to the client, with added real-time streaming capabilities.

  • MCP Endpoints: Provides /mcp/list_tools and /mcp/call_tool endpoints that mimic MCP tool discovery and invocation, with expanded protocol support including prompts, resources, and roots.

  • Real-time Streaming: Server-Sent Events (SSE) for live command execution updates and terminal output.

  • Security Enhancements: Basic authentication, rate limiting, input validation, and security headers for safer operation.

  • Multiple Transport Options: Supports both HTTP REST API and STDIO command-line interface for flexibility.

  • Session Management: Improved session handling and conversation storage.

  • Docker Support: Containerization for easier deployment and testing.

Getting Started

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 the DEEPSEEK_API_KEY in .env with your DeepSeek API key.

Running the Server

Quick Start

Use the provided startup script for convenience:

chmod +x startup.sh
./startup.sh

Manual Start

Run the Flask server with:

python server.py

Visit http://127.0.0.1:8000 to access the chat interface.

Docker

For containerized deployment:

docker-compose up -d

Endpoints

Chat Endpoint

  • URL: /chat
  • Method: POST
  • Payload: { "message": "your message here" }
  • Description: Adds the user message to the conversation, sends it to the DeepSeek API, looks for any command instructions (CMD:), executes them, and returns the final response.

MCP Endpoints

List Tools

  • URL: /mcp/list_tools
  • Method: POST
  • Response: JSON listing 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: Directly invoke a tool command on the server.

Real-time Streaming

  • URL: /stream?session_id=session_id
  • Method: GET
  • Description: Server-Sent Events endpoint for real-time command output updates.

Future Improvements

  • Protocol Standardization: Implement JSON‑RPC for a more robust and standardized communication protocol.

  • Real-time Communication: Add Server‑Sent Events (SSE) or WebSockets for live command output streaming.

  • Session & Security Enhancements: Introduce per‑user sessions, proper authentication, input sanitization, and comprehensive error handling.

  • Modular Code Architecture: Further separate API logic from business logic for better maintainability and scalability.

License

This project is open-source and available under the MIT License.

Ver o original no GitHub

Bom para

Relacionados