1mcpserver

🚀 1 MCP Server: A MCP server that picks and configures MCP servers for you

This is an MCP server that can automatically discover and configure MCP servers on your local machine. It's a remote solution that aims to simplify your MCP server experience.

Our goal is to provide just this one MCP server. You can then leave all the other tasks, such as searching for servers, selecting servers, and configuring servers, to this MCP server. There's no need to run setup commands or acquire API keys. You only need to modify one file.

Demo video: Click here

🚀 Quick Start

✨ Features

• Automatic discovery and configuration of MCP servers.
• Supports both remote and local setups.
• Simplifies the process of using MCP servers.

📦 Installation

Option 1: Simple remote setup - integration with Cursor and Claude

Add the following to the Cursor or Claude MCP config file:

• For Cursor: Open ~/.cursor/mcp.json.
• For Claude:
  • On macOS, open ~/Library/Application\ Support/Claude/claude_desktop_config.json.
  • On Windows, open %APPDATA%\Claude\claude_desktop_config.json.

Add the following to the file:

{
  "mcpServers": {
    "mcp-server-discovery": {
      "url": "https://mcp.1mcpserver.com/mcp/",
      "headers": {
        "Accept": "text/event-stream",
        "Cache-Control": "no-cache",
        "API_KEY": "value"
      }
    }
  }
}

If you're already using other servers, the JSON file should look like this:

{
    "mcpServers": {
        "mcp-server-discovery": {
            "url": "https://mcp.1mcpserver.com/mcp/",
            "headers": {
                "Accept": "text/event-stream",
                "Cache-Control": "no-cache"
            }
        },
        "file-system": {
            "command": "node",
            "args": [
                "/Users/jiazhenghao/CodingProjects/MCP/filesystem/index.ts",
                "~/"
            ]
        }
    }
}

Option 2: Local Setup with STDIO

git clone https://github.com/particlefuture/MCPDiscovery.git
cd MCPDiscovery
uv sync
uv run server.py

Unfortunately, as of the time this README was updated, Claude only allows STDIO. So you'll have to modify server.py to use STDIO. Find the main block in server.py, comment out the "Streamable HTTP server BLOCK" and uncomment the "Standard I/O server BLOCK". The final code should look like this:

    asyncio.run(
        mcp.run_async(
            transport="stdio",
        )
    )

The mcp.json should look like this:

{
    "mcpServers": {
        "mcp-servers-discovery": {
            "command": "/Users/jiazhenghao/.local/bin/uv",
            "args": [
                "--directory",
                "{PATH_TO_THE_CLONED_REPO}",
                "run",
                "server.py", 
               "--local"
            ]
        },
        "file-system": {
            "command": "node",
            "args": [
                "{FILE_SYSTEM_CLONED_PATH}/filesystem/index.ts",
                "~/"
            ]
        }
    }
}

💻 Usage Examples

Basic Usage

After setting up the MCP server according to the above steps, you can start using it with Cursor or Claude. Simply make requests to the configured MCP server, and it will handle the rest.

📚 Documentation

Architecture

There are two types of search tools:

• Quick Search: When the user has an explicit goal of what type of MCP they want (e.g., "I want a MCP server that handles payment"), this tool returns a list of MCP servers.
• Deep Search ^*: When the user has a high-level or complex description of the goal (e.g., "Build me a website that analyzes other websites"). The LLM needs to break it down into multiple steps and components (e.g., "I need to analyze the website traffic", "I need to analyze the website tech stack", "I need to show some web data", etc.), then find MCP servers for each step. If a corresponding MCP server doesn't exist, the user will be informed to decide whether to ignore this component, break it down further, or implement it themselves.

There are multiple stages in the deep search:

1. Planning stage:
   • Set up MCP servers:
     • Get and configure API keys as needed, and provide users with instructions on obtaining API keys.
     • Modify the mcp.json files.
2. Testing stage:
   • Test to see if the servers are working. Call the test_server_template_code tool, which returns a simple client testing code example.
3. Acting stage:
   • Build the workflow/application by calling the MCP servers.

*We're supposed to put deep search as a prompt, but both Cursor and Claude rarely call prompts.

🔧 Technical Details

The deep search process involves horizontal and vertical expansion. Horizontal expansion is for finding independent components, while vertical expansion is for finding steps that have to be done sequentially (e.g., fetch, analyze, generate graph).

📄 License

Not provided in the original document.

📈 Change Log

• July 31, 2025: Upgrade to 0.2.0. Added agentic planning. For complex tasks, the server now prompts the LLM to perform multi-step MCP server query.

🚀 Future Plans

• Improve the demo videos: new domain name, actual example, voice explanation.
• Call For MCCP (Model Context Communication Protocol): A standard way of communicating between MCP servers. Motivation: Allow directly sending requests to other MCP servers (each MCP server might also have dependencies). (But would also need stricter supervision).
• Avoid calling functions with a leading prefix internal_ unless instructed by MCP servers.
• Create a better database for MCP servers. It should be in the structure: server, description, url, config json, (optionally, additional setup, docker, api_key, etc).

📚 References

This repo is based on these repos. Huge thanks to the authors and contributors of these repos:

• wong2/awesome-mcp-servers
• metorial/mcp-containers
• punkpeye/awesome-mcp-servers
• modelcontextprotocol/servers

This project is published to:

• https://mcpservers.org/
• https://glama.ai/mcp/servers

📋 More about the repo

Motivation and Context

Currently, most MCP server searches are done via GitHub searching or online Google search. There have been several MCP wrappers, but most serve as middleware infrastructure that hosts different MCPs at an endpoint. The gathering of which MCP to host is still mostly done by manual search. We provide an automatic pipeline for not only searching but also automatically configuring MCP servers.

How Has This Been Tested?

Tested with Claude and Cursor using both remote endpoints and local STDIO. The demo has been tested 10 times.

⚠️ Troubleshooting

• If using a virtual environment and encountering a ModuleNotFoundError even after installing the module, delete the virtual environment and create a new one.