Pydantic AI 2.0: Composable Agent Capabilities
Cole Medin · youtube · 2026-07-10
The video introduces the new approach of Pydantic AI 2.0: designing agents as composed of multiple capabilities, rather than lumping tools, prompts, and guardrails into a monolith.
The author emphasizes that capabilities feature:
- Their own instructions, tools, model settings, hooks
- The ability to be combined like building blocks into different agents
- Reusability across multiple agents without code changes
The video also demos a real-world support agent and reuses one of its capabilities in another agent, demonstrating how "composable capabilities" make production-grade agents easier to maintain. Resource links mention:
- Example project orbit-support-agent
- pydantic-ai GitHub repository
- Official Pydantic AI 2.0 article
The overall focus isn't just on models, but on how to structure agents to be reusable and scalable.
More from coding & agent
- Soft Clamp cuts tool-call overuse in multi-teacher distillation, from 13.7% to 9.0% — antgroup · 2026-07-21
- Agent harness memory loss and compaction are still a major usability problem — adityaag · 2026-07-21
- SpecJudge runs locally on Ollama to pick the right-sized AI model for your project — jokiruiz · 2026-07-21
- A developer maps out six design rules for CLIs that humans and AI agents can both use — yujiezha · 2026-07-21
- A coding-agent skill that forces ADHD-friendly, answer-first output — ayghri · 2026-07-21
- A set of agent skills for CAD, robotics, and hardware design — earthtojake · 2026-07-21