Tutorial Weathermcp

🚀 MCP Tutorial Overview

📚 References

• Reference Link

• [Completed] Try using the MCP server provided in the tutorial via cline.

  • Error Resolution: When cline outputs debug-level logs, it was assumed that errors were also being sent to stderr. To resolve this issue, the loglevel was set to error (Link).
  • Tool Explanation: There is an issue where the argument descriptions in Tools do not display correctly. It is necessary to check if this is a bug.

• [Completed] Reinstall the Python Lint Checker and Formatter (-> Installed with ruff).

Note: API requests are not used. During the process where the LLM views and judges (my requests + the tools provided by MCP servers) and then executes them automatically, up to two API requests may occur. The costs are estimated to be around $0.05 and $0.03. In any case, the tools can be effectively used to obtain the necessary data.

➕ Example Addition

Refer to the following link to understand the content of the Python SDK: Python SDK GitHub Link

1. Resource Definition

This is the definition of the resources that the MCP server intends to provide to the MCP client.

• Resource Discovery (How does the MCP client discover the resources available on the MCP server?)
  • Direct Resources: The resources are exposed in the following format via the resources/list endpoint.

{
  uri: string;  // Unique identifier of the resource
  name: string; // Human-readable resource name
  description?: string; // Optional description
  mimeType?: string; // Optional MIME type
}

• Resource Templates: The resources provided via uritemplate are exposed.

{
  uriTemplate: string;   // URI template according to RFC 6570
  name: string;          // Human-readable name of this type
  description?: string;  // Optional description
  mimeType?: string;     // Optional MIME type for all matching resources
}

• If the MCP client discovers the resources available on the MCP server through resource discovery, how can it call them?
  • The method to define a resource red handler is as follows:
    • Define a function that accepts a uri as an argument using the @mcp.resource() decorator.
    • Or define it via @mcp.resource(uri_template).

Summary: The MCP server either directly provides the URIs of the resources or exposes them through templates to enable resource discovery. After the MCP client discovers the available resources, it can access the resources via the corresponding URIs.

2. Prompts Definition

Standardized methods are provided during LLM interactions by offering prompts tailored to the purpose. Users can explicitly select the prompts.

APIs to be implemented:

• Prompt Discovery
  • Return value: (name, description, arguments)
• Prompt Usage
  • Arguments: (name, arguments declared in prompts)
  • Return value: Prompts (supports dynamic (resource-based prompts) and multiple prompts)

3. Image Return

There is a function to return images.

Note: By leveraging the various functions and APIs provided by MCP, workflows can be built, and an environment can be created where the LLM can analyze and process data.