MCP Proxy Pydantic Agent

🚀 Example: Integrating MCP (Model Context Protocol) Server with Pydantic.AI

This example uses some content from https://github.com/modelcontextprotocol/quickstart-resources.git, especially the weather "server" code.

Two different LLMs (Large Language Models) are used in the code just for demonstration purposes. The agent tool uses gpt-4o, while the tool itself uses sonnet. Therefore, please export OPENAI_API_KEY and ANTHROPIC_API_KEY - or modify the code to suit your models.

The pyproject.toml assumes you are using the 'uv' package manager.

🚀 Quick Start

Steps to Run

1. Clone this repository.
2. Execute uv sync.
3. Switch to the mcp-client directory.
4. Execute uv run client.py (This requires OpenAI and Anthropic keys and directly uses the Anthropic library).
5. Alternatively, execute uv run client2.py (This version only uses PydanticAI and is suitable for any function-calling LLM).

(Or try running client2.py - this version only uses Pydantic and any function that can call an LLM).

Now, try asking some questions, for example:

What time is it in Bangalore when it's 7:30 PM in New York?

What's the current weather in Chicago? (Exit after completion)