# PydanticAI (pydantic-ai) - AgentBreakWatch Reference

> TL;DR Verdict: PydanticAI (`pydantic-ai`) is currently at v2.61.0 on PYPI (published 2026-10-24), with a 90-day Stability Score of 70/100 (C+) and HIGH schema drift risk across 56 releases in the last 90 days. v2.x replaces result_type with output_type on Agent(), and renames result.data to result.output across all run methods.

## Live Telemetry Metadata
- Canonical Page: https://agentbreakwatch.vercel.app/p/pydantic-ai
- Category Hub: https://agentbreakwatch.vercel.app/c/agent-sdks (AI Agent SDKs & Frameworks Breaking-Change Index)
- Package Name: `pydantic-ai`
- Registry: PYPI (https://pypi.org/project/pydantic-ai/)
- GitHub Repository: https://github.com/pydantic/pydantic-ai
- Vendor: Pydantic
- Latest Live Version: v2.61.0
- Published Date: 2026-10-24T03:55:19.172362Z
- 90-Day Release Count: 56
- Direct Dependencies: 31
- License: MIT
- Stability Score: 70/100 (Grade C+)
- Schema Drift Risk: HIGH
- Recommended Exact Pin Command: `pip install "pydantic-ai==2.61.0"`
- Watched Symbols / MCP Tools: Agent, RunContext, MCPServerStdio, MCPServerStreamableHTTP
- Snapshot Timestamp: 2026-10-24T12:11:27.697Z

## Breaking Change & Schema Drift Summary
v2.x replaces result_type with output_type on Agent(), and renames result.data to result.output across all run methods.

## Migration Diff: Before (Deprecated / Unpinned Pattern)
```
agent = Agent("openai:gpt-4o", result_type=MyModel)
result = await agent.run("hello")
print(result.data)
```

## Migration Diff: After (Verified v2.61.0 Pattern)
```
agent = Agent("openai:gpt-5", output_type=MyModel)
result = await agent.run("hello")
print(result.output)
```

## Recent Verified Releases
- v2.61.0 (published 2026-10-24) [LATEST]
- v2.60.0 (published 2026-10-23)
- v2.59.0 (published 2026-10-21)
- v1.107.9 (published 2026-10-21)
- v1.107.8 (published 2026-10-20)
- v2.58.0 (published 2026-10-20)