Simple MCP Tool Server

A simple MCP server that exposes a website fetching tool using SSE transport.

MCP Python SDK Documentation

The MCP Python SDK documentation has been split into smaller files and organized in the docs/ directory. This structure makes it easier for AI agents to navigate and understand the SDK. The documentation covers:

• Core concepts (servers, resources, tools, etc.)
• Running MCP servers in different modes
• Examples and advanced usage
• And more!

Server Tools

The server exposes the following tools:

• fetch: Remote HTTP fetcher – give an absolute URL; returns page text.

  • url: The URL of the website to fetch (required)

• search_docs: Semantic search across SDK documentation; returns top-k excerpts.

  • query: Search phrase or question (required)
  • k: Number of top matches to return (optional, default = 3)

• get_content: Get the full local file for any match returned by search_docs.

  • file: Path relative to docs (required)

Project Architecture

mcp_simple_tool/
    __init__.py          # Package initialization
    __main__.py          # Entry point when run as module
    cli.py               # Command-line interface
    server/              # Server implementation
        __init__.py      # Server package initialization
        app.py           # ASGI application setup
        config.py        # Configuration settings
        handlers.py      # Tool implementations
        http.py          # HTTP utilities
    semantic_search/     # Semantic search functionality
        __init__.py      # Package initialization
        indexing.py      # Build and persist vector store
        search.py        # Load index and query helpers

Using with Cursor

This MCP server can be used with Cursor as a client. For setup:

1. Run the server in a terminal:

source venv/bin/activate
python -m mcp_simple_tool start
# or use the restart command
python -m mcp_simple_tool restart

2. Configure Cursor by creating a .cursor/mcp.json file:

{
  "mcpServers": {
    "website-fetcher-sse": {
      "url": "http://localhost:7000/sse"
    }
  }
}

3. Mention the server in your prompts when using Cursor

Requirements

• Python 3.10 or higher (tested on Python 3.13)

Installation

# Create a virtual environment
python3 -m venv venv

# Activate the virtual environment
source venv/bin/activate

# Install the package and dependencies
pip install -r requirements.txt

Usage

The package provides a command-line interface (CLI) with several commands to manage the MCP server:

Starting the server

Start the server on the default port (7000) or specify a custom port:

# Using default port (7000)
python -m mcp_simple_tool start

# Using custom port
python -m mcp_simple_tool start --port 8000

Managing the server

# Check if server is running
python -m mcp_simple_tool check [--port PORT]

# Stop the server
python -m mcp_simple_tool stop [--port PORT]

# Restart the server (stop and start)
python -m mcp_simple_tool restart [--port PORT]

The restart command will:

1. Stop any existing server on the specified port
2. Start a new server in the background
3. Wait until the server is responsive
4. Log output to server.log

CLI quick reference

Development Setup

For development, install additional tools:

pip install -e .
pip install -r requirements.txt

Use the Makefile for common tasks:

# Format code
make fmt

# Run linters
make lint

# Run tests
make test

The test suite has a built-in 20-second timeout for all tests to prevent hanging, especially with SSE endpoints. For individual tests, a more strict timeout can be specified using the @pytest.mark.timeout(seconds) decorator.

Semantic Search Index

For the search_docs tool, you can manually build or rebuild the vector index:

# Build or rebuild the semantic search index
python scripts/build_doc_index.py

The index is built automatically on first tool use if it doesn't exist.