This MCP server is designed to collect MCP servers from the internet and provide tools to submit them to the MCP Server Directory. The server exposes three tools: extract-mcp-servers-from-url, which parses a given URL to discover MCP servers; extract-mcp-servers-from-content, which analyzes raw content to extract MCP servers; and submit-mcp-server, which sends a discovered MCP server to the MCP Server Directory (like mcp.so) with an optional avatar URL. To run the collector locally, use the UV-based runner configuration and provide your OpenAI credentials and the MCP submit endpoint. This setup allows automated extraction and submission workflows powered by AI-assisted analysis and prompts.

Prerequisites:

• Python (recommended 3.11+)
• UV CLI installed (used to run MCP servers, e.g., uv or uvx)
• Network access to the MCP Server Directory submission endpoint

Installation steps:

1. Install Python and ensure python --version and pip --version work.

2. Install the UV CLI (the tool used to run MCP servers). Example (adjust to your environment):

   If using pipx (recommended)

   pipx install uv

   If using pip directly (alternative)

   pip install uv

3. Clone or download this MCP server repository (mcp-server-collector).

4. Install any Python dependencies required by the collector (if a requirements.txt is provided):

   pip install -r requirements.txt

5. Configure environment variables (see additional_notes for details). You can set them in your shell or a .env file as needed.

6. Run the collector using the configured command (example shown in mcp_config):

   uvx mcp-server-collector

7. Verify the server starts correctly and check logs for any errors. If you are using the integrated inspector, you can attach debugging tooling as described in the repository README.

Note: The exact installation commands may vary depending on how UV is installed in your environment. Refer to the UV documentation for your platform if the above commands differ.

Environment variables and configuration tips:

• OPENAI_API_KEY, OPENAI_BASE_URL, and OPENAI_MODEL are used to configure AI-assisted tooling within the collector. Ensure these are set to valid credentials and endpoints.
• MCP_SERVER_SUBMIT_URL is the endpoint where discovered MCP servers are submitted to the central directory (e.g., mcp.so). Verify the URL is reachable and accepts the required payload.
• If you run the server in development mode, you may switch between different runners (uv or uvx) depending on how you manage dependencies and directory structure.
• When debugging, use the MCP Inspector or similar tooling to visualize prompts, tool calls, and responses for easier troubleshooting.
• If you encounter issues with environment variable resolution, consider creating a dedicated .env file and loading it with your shell or the UV runner configuration.
• The collector supports three tools; ensure the inputs (url or content) are valid strings and handle errors gracefully in your integration layer.