MCP Research Router

🚀 MCP Research Router

MCP Research Router is an aggregator and intelligent router that simplifies access to multiple MCP servers, recommends suitable tools, and boosts performance through batch parallel execution.

🚀 Quick Start

1. Installation

npm install -g mcp-research-router
# Or run directly
npx mcp-research-router

2. Configuration

Recommendation: Use environment variables

Add environment variables in the MCP client:

MCP_LLM_ENABLED=true
MCP_LLM_API_KEY=your-api-key-here
MCP_LLM_BASE_URL=https://open.bigmodel.cn/api/paas/v4
MCP_LLM_MODEL=glm-4.7-flash
MCP_SERVER_URL=http://127.0.0.1:3000/mcp/your-group-id

3. Connect to the client

Edit the configuration file (for clients that support MCP, such as Claude Desktop and Cursor):

• Claude Desktop:
  • Windows: %APPDATA%\Claude\claude_desktop_config.json
  • Mac: ~/Library/Application Support/Claude/claude_desktop_config.json
• Cursor:
  • Windows: %APPDATA%\Cursor\User\globalStorage\mcp_settings.json
  • Mac: ~/Library/Application Support/Cursor/User/globalStorage/mcp_settings.json

Minimum configuration (only connect to the service):

{
  "mcpServers": {
    "mcp-research-router": {
      "command": "npx",
      "args": ["mcp-research-router"]
    }
  }
}

Including environment variables:

{
  "mcpServers": {
    "mcp-research-router": {
      "command": "npx",
      "args": ["mcp-research-router"],
      "env": {
        "MCP_LLM_ENABLED": "true",
        "MCP_LLM_BASE_URL": "https://ark.cn-beijing.volces.com/api/coding/v3",
        "MCP_LLM_API_KEY": "your-api-key-here",
        "MCP_LLM_MODEL": "ark-code-latest",
        "MCP_SERVER_ENABLED": "true",
        "MCP_SERVER_URL": "http://your-server-url/mcp/endpoint",
        "MCP_SERVER_NAME": "my-server",
        "MCP_SERVER_HEADERS": "{\"Authorization\": \"Bearer your-token\"}"
      }
    }
  }
}

✨ Features

1. Unified management of multiple MCP servers

Aggregate multiple MCP servers, allowing you to access all tools with a single connection. There's no need to add each server to the client one by one.

2. Intelligent tool recommendation

Based on your needs, the system automatically recommends the most suitable tools through LLM. You no longer need to browse through long tool lists.

3. Batch parallel execution

Call multiple relevant tools simultaneously, improving performance by 3 - 5 times. This is suitable for scenarios that require comprehensive results from multiple tools.

💻 Usage Examples

Get the tool list

{
  "name": "get_tool_list"
}

Intelligent tool recommendation

{
  "name": "get_tool_list",
  "arguments": {
    "user_query": "Search for the latest news about artificial intelligence",
    "max_tools": 3
  }
}

The system will recommend the most suitable tools based on your needs.

Batch execute tools

{
  "name": "execute_tools",
  "arguments": {
    "tools": [
      {
        "tool_name": "server-name-metaso-metaso_web_search",
        "arguments": {"q": "Latest developments in artificial intelligence"}
      }
    ]
  }
}

📚 Documentation

Is this tool suitable for you?

✅ Suitable for use

• You need to manage multiple MCP servers centrally.
• You have a large number of tools and don't know which one to choose.
• You need to call multiple tools in batches.
• You want to use LLM for intelligent tool recommendation.

❌ Not suitable for use

• You only have one MCP server.
• You have a small number of tools and can choose directly.
• You don't need the intelligent recommendation function.

If your client already supports adding multiple MCPs and you have only a few tools, you don't need this tool.

Configuration instructions

Environment variables (recommended)

How to use MCP_SERVER_HEADERS

You need to fill in a JSON-formatted string:

MCP_SERVER_HEADERS={"Authorization": "Bearer your-token"}

Examples:

• Bearer Token authentication: {"Authorization": "Bearer your-token"}
• API Key authentication: {"X-API-Key": "your-api-key"}
• Multiple request headers: {"Authorization": "Bearer your-token", "X-Custom-Header": "value"}

Configuration file method

If you need to use a configuration file, edit config.json:

{
  "subAgent": {
    "enabled": true,
    "type": "openai",
    "baseUrl": "https://open.bigmodel.cn/api/paas/v4",
    "apiKey": "your-api-key-here",
    "model": "glm-4.7-flash",
    "timeout": 90000
  },
  "mcpServers": {
    "enabled": true,
    "servers": [
      {
        "name": "my-favorites",
        "type": "streamable-http",
        "url": "http://127.0.0.1:3000/mcp/your-group-id"
      }
    ]
  }
}

For a complete configuration example, please refer to config.full.example.json.

Prompt modes

The system supports multiple prompt modes, and you can choose according to your needs:

Predefined modes

• tool_recommendation (default): Quick tool recommendation
• deep_research: Deep research mode, suitable for complex tasks

Custom mode

Create a new folder in the prompts/ directory:

prompts/
└── my_mode/
    ├── system.txt
    └── user.txt

Pass prompt_mode: "my_mode" when using it.

For detailed instructions, please refer to prompts/README.md.

Frequently Asked Questions

Q: How to configure MCPHub groups?

1. Create a group on the MCPHub web page.
2. Add commonly used MCP servers to the group.
3. Copy the group ID.
4. Configure the environment variable: MCP_SERVER_URL=http://127.0.0.1:3000/mcp/group-ID

⚠️ Do not add MCP Research Router itself to the group, otherwise, it will cause an infinite loop.

Q: Can I use it without configuring LLM? Yes, but the functionality is limited:

• ✅ You can get the tool list.
• ✅ You can execute tools.
• ❌ You cannot use the intelligent recommendation function.

🔧 Technical Details

# Install dependencies
pnpm install

# Development mode (supports hot reloading)
pnpm dev

# Build
pnpm build

# Run
pnpm start

📄 License

MIT

🔗 Related Links

• GitHub Repository
• Installation Guide
• Deployment Documentation
• Changelog