Skip to content

Pydantic AI 2.61 adds structured citations and durable extensions

Updated Pydantic AI

Summary

Pydantic AI adds structured text citations, provider normalization and a public durable-engine extension surface, with compatibility changes.

Category
Frameworks
Date
Pricing
Not stated in the sources we cited.

Pydantic AI published v2.61.0 on October 24, 2026.

Citations across the application

The release adds structured citations to TextPart, normalizes citations from Bedrock, OpenRouter and xAI, and replays citations in message history to the provider that produced them. The Vercel AI adapter also preserves citations.

That is useful for applications that need citation data to survive transport between model responses, stored conversation history and a user interface. It does not by itself verify whether a cited source supports a model's answer.

Extension and compatibility changes

The durable-engine extension surface becomes public, including operation types, parameter transports, run context and run-identity hooks. Hosted MCP capabilities accept asynchronous authentication functions, while shell and background-job logs gain a size bound.

The compatibility notes are substantial. They cover thinking-setting precedence, models that cannot disable thinking, unsupported file returns and guardrail behavior. Treat this as an upgrade requiring review of the capabilities your application uses, not a purely additive patch.

See the Pydantic AI hub and the full official release notes for affected interfaces.

More from Pydantic AI

View all

Questions

What is Pydantic AI?

The product site describes it as a model-agnostic Python framework featuring type safety, structured outputs, validation and observability.

Is Pydantic AI open source?

The GitHub repository reports an MIT license.

Which models does it support?

The product site says it supports OpenAI, Anthropic, Google and more.

Filed on Agent Radar. Claims in this note are limited to the sources above.