MCPDataAnalytics

# Lesson 002 - Your First MCP Server **Objective:** Create an MCP server that can communicate with a host application. **Outcome:** Create an MCP server in Python that will say "Hello World". ### Steps 1. Understand what imports are required. 2. Initialize the server. 3. List the tools: - Understand tool names - Understand descriptions - Understand the input schema 4. Set up tool calling. 5. Understand how `FastMCP` is more concise than `mcp`. 6. Set up the config and test the server in VS Code and Claude Desktop. ### Key takeaways - A server can be as simple as a single Python script. - More advanced servers will be projects of multiple files. - A server has several common components: - (Optional) Docstring for documenting the script / file / server. - Imports for necessary libraries - Server instance - Tool registry - Listing tools - Calling tools - Each tool must have: - A name. - Any arguments and their explicit types. - Output types. - (Optional) Description, which is necessary for the LLM to "find" the tool effectively . - Runner (to run the server) - Suggestions for code hygeine: - Format your code. - Use error handling. - Organize it into "code regions" - In Python, the `FastMCP` library takes care of the basic boilerplate. - This helps make your code more concise. - It is good to try a few times with the `mcp` library first to understand it.