Mcp_proxy_pydantic_agent

What is Mcp_proxy_pydantic_agent?

Mcp_proxy_pydantic_agent is a sample integration that shows how to connect Model Context Protocol (MCP) servers with Pydantic.AI. It demonstrates using a weather server example and is intended for developers who want to mix MCP tools with PydanticAI agents.

How to use Mcp_proxy_pydantic_agent?

Clone the repo, run uv sync, change into mcp-client, then run either client.py (requires OpenAI and Anthropic API keys and uses Anthropic libraries directly) or client2.py (uses only PydanticAI and works with any function‑calling LLM). Export OPENAI_API_KEY and ANTHROPIC_API_KEY before running.

Key features of Mcp_proxy_pydantic_agent

• Integrates MCP servers with Pydantic.AI agents
• Uses two different LLMs (gpt‑4o and sonnet) for demonstration
• Pure PydanticAI option via client2.py (no Anthropic library dependency)
• Works with any function‑calling LLM in the pure Pydantic variant

Use cases of Mcp_proxy_pydantic_agent

• Ask for the time in New York when given a time in another city
• Ask for the current weather in a specified location
• General demonstration of combining MCP tools with PydanticAI

FAQ from Mcp_proxy_pydantic_agent

What dependencies are required?

—

Do I need both OpenAI and Anthropic API keys?

The proxy agent uses gpt‑4o and the tool uses sonnet. You must export both OPENAI_API_KEY and ANTHROPIC_API_KEY, unless you modify the code to use different models.

What is the difference between client.py and client2.py?

client.py uses Anthropic libraries directly. client2.py uses only PydanticAI and works with any function‑calling LLM.

What package manager is required?

The project uses the uv package manager (as configured in pyproject.toml).

Where does the weather server code come from?

The weather server code is adapted from the MCP quickstart resources repository.