Industry story
Pydantic AI v2.19.0 adds HTTP error metadata and FastMCP 4 support
agent-framework mcp reliability tool-use
Pydantic AI v2.19.0 is a small release that solves a real operational annoyance. Before this, handling rate-limit responses in agents meant writing custom parsing logic per provider; now ModelHTTPError surfaces headers and retry_after directly, and every provider SDK populates them automatically. The FastMCP dependency range also widens to version 4, so you're not stuck pinning MCP tooling to stay compatible. Not a headline drop, but the kind of plumbing fix that quietly removes friction from production agent deployments.
Full analysis
Pydantic AI is a Python agent framework built around Pydantic's data validation. It released v2.19.0 on July 27, 2026. The main feature adds headers and retry_after fields to ModelHTTPError, populated automatically across all provider SDKs, making it easier for agents to handle rate-limit responses gracefully without custom parsing. The release also widens the FastMCP (a Model Context Protocol server library) dependency range to support up to version 4, keeping compatibility with the latest MCP tooling.
Comments