MCP Testing Project

This project demonstrates a simple client-server setup using the fastmcp library for Model Context Protocol (MCP) communication. It includes:

• A server that exposes tools and resources via MCP (see server/server.py and server/mcp_blog_server.py)

• A client that connects to the server, lists available tools, and calls a sample tool (see client/client.py)

Advanced Example: Blog MCP Server

The file server/mcp_blog_server.py demonstrates advanced MCP server features:

• Dynamic Resources: Exposes blog posts, single post, and category-based resources with caching and templates.

• Complex Tools: Tools for creating posts, publishing workflows, batch tag updates, and searching posts with parameter validation and caching.

• Caching: Uses cachetools.TTLCache for efficient response caching.

• Tool Chaining: Implements multi-step workflows (e.g., publishing a post and notifying subscribers).

• Asynchronous Operations: Batch operations and async processing for scalability.

To run the advanced blog server:

uv run fastmcp run server/mcp_blog_server.py

You can then connect with a compatible MCP client to interact with the blog tools and resources.

Related MCP server: FastMCP Training Course Server

Project Structure

MCP_testing/
│
├── main.py
├── pyproject.toml
├── README.md
├── server.json
├── client/
│   └── client.py
├── server/
│   ├── server.py
│   └── mcp_blog_server.py
└── ...

Requirements

• Python 3.8+

• fastmcp

• uv (for fast Python package management and running)

Install dependencies:

uv pip install fastmcp

Running the Server

From the project root, run:

uv run fastmcp run server/server.py

This will start the MCP server. Make sure the server is running before starting the client.

Running the Client

The client reads the MCP server URL from server.json and connects to it, listing available tools and calling the get_weather tool as an example.

From the client directory, run:

uv run client.py

Configuration

The server.json file should look like this:

{
  "mcpServers": {
    "weather-mcp": {
      "url": "http://127.0.0.1:8000/sse",
      "transport": "sse"
    }
  }
}

Example Output

✅ Connected to MCP server!
  • get_weather: Get the weather for a city
Result: { ... }

Notes

• The client and server are both written in Python and use the fastmcp library for communication.

• You can add more tools to the server and call them from the client as needed.

For more information, see the fastmcp documentation or the code in client/client.py, server/server.py, and server/mcp_blog_server.py.