🧠 Expert MCP

Equip your AI model with an "on-call" senior expert consultant.

When downstream models encounter complex problems, they can use this MCP tool to forward the query to a pre-configured advanced model (e.g., GPT-5.5, Claude Opus 4.7, etc.) to obtain deep analysis and professional advice, then combine it with their own judgment to provide a final answer.

✨ Features

• 🔌 OpenAI Compatible — Connects to any upstream endpoint in OpenAI format (OpenAI / DeepSeek / Qwen / vLLM / Ollama, etc.)

• ⚙️ Config-Driven — All parameters are centrally managed in config.json, no code changes required

• 📡 Streamable HTTP — Compliant with the latest MCP standard for streaming HTTP transport, endpoint /mcp

• 🛠️ Rich Tool Prompts — Carefully designed description to guide downstream models to call at the right time

• 🧩 Three-Part Input — question (required), context (background), focus (key areas)

• 📊 Full Logging — Request logs + Token usage statistics for easy monitoring and troubleshooting

Related MCP server: Consult LLM MCP

🏗️ How It Works

┌─────────────────────────────────────────────────────────┐
│                      用户 (User)                         │
└───────────────────────────┬─────────────────────────────┘
                            │ 提问
                            ▼
┌─────────────────────────────────────────────────────────┐
│              下游模型 (Claude / GPT / Qwen …)            │
│                                                         │
│   遇到复杂问题？→ 调用 consult_advanced_model 工具       │
└───────────────────────────┬─────────────────────────────┘
                            │ MCP Streamable HTTP
                            ▼
┌─────────────────────────────────────────────────────────┐
│                   Expert MCP Server                     │
│               (本项目 server.py)                        │
└───────────────────────────┬─────────────────────────────┘
                            │ OpenAI API 请求
                            ▼
┌─────────────────────────────────────────────────────────┐
│         上游高级模型 (GPT-5.5、Claude Opus 4.7 …)      │
│                   返回深度分析意见                       │
└─────────────────────────────────────────────────────────┘

📦 Quick Start

1. Clone the repository

git clone https://github.com/MineJPGcraft/Expert-mcp.git
cd Expert-mcp

2. Install dependencies

pip install -r requirements.txt

Python 3.10+ is recommended; installation within a virtual environment is suggested.

3. Edit the configuration file

Copy and modify config.json:

{
  "host": "0.0.0.0",
  "port": 8765,
  "upstream": {
    "base_url": "https://api.openai.com/v1",
    "api_key": "sk-xxxxxxxxxxxxxxxxxxxx",
    "model": "gpt-5.5",
    "temperature": 0.3,
    "max_tokens": 4096,
    "timeout": 120,
    "system_prompt": "你是一位顶尖的资深专家顾问，请对问题进行深入、严谨、可执行的分析。"
  }
}

See ⚙️ Configuration Reference below for details on configuration items.

4. Start the service

python server.py

Successful startup is indicated by the following logs:

2025-xx-xx | INFO    | mcp-advisor | 上游模型: gpt-4o @ https://api.openai.com/v1
2025-xx-xx | INFO    | mcp-advisor | MCP 监听: http://0.0.0.0:8765/mcp

MCP endpoint address:

http://127.0.0.1:8765/mcp

🔧 Client Integration

Add the following configuration in clients that support MCP Streamable HTTP (Cherry Studio, Cline, Claude Code, etc.):

{
  "mcpServers": {
    "expert-advisor": {
      "type": "streamableHttp",
      "url": "http://127.0.0.1:8765/mcp"
    }
  }
}

Remote deployment? Replace 127.0.0.1 with the actual IP or domain name of the server and ensure the corresponding port is open in the firewall.

⚙️ Configuration Reference

Switch configuration files using environment variables

MCP_CONFIG=config.prod.json python server.py

🛠️ Tool Description

Tools available for downstream models to call:

consult_advanced_model

Recommended scenarios for calling:

• Problems where your own confidence is < 80%

• User explicitly requests deep thinking / rigorous analysis / best practices

• Multi-constraint problems involving complex trade-offs

• Mathematics, algorithms, system design, or difficult bugs requiring step-by-step reasoning

• Situations where you need to verify your own conclusions

Scenarios not recommended:

• Simple greetings or pure information lookups

• Basic questions with obvious answers

• High-frequency, repetitive simple tasks

🌐 Connecting to Upstream Services

Simply modify the upstream section in config.json to connect to different service providers.

📋 Dependencies

mcp>=1.2.0
openai>=1.40.0

📄 License

MIT License © 2026 MCJPG