Pydantic AI
A linear, one-concept-per-lesson path through Pydantic AI: a type-safe agent framework from the Pydantic team, where an agent's output is a validated Pydantic model, not a string you hope is JSON. 24 lessons · 3 tiers.
Prerequisites: Completion of LangChain & Agent Building, or already comfortable with @tool, bind_tools, and with_structured_output. Every lesson calls out the LangChain equivalent of the concept it introduces. Lesson 17 also assumes LangGraph Lessons 1-6, Lesson 18 assumes LangSmith Lessons 1-4, and Lesson 21 assumes the MCP course through Lesson 12.
Lessons use Google's Gemini free tier via the model string "google:gemini-3.5-flash-lite", Pydantic AI's equivalent of the ChatGoogleGenerativeAI model object in the LangChain course. A free GOOGLE_API_KEY from aistudio.google.com/apikey is all that's needed.
Built on Pydantic AI, the framework this course teaches.
Beginner
The core Agent API: typed output, system prompts, dependency injection, tools, and streaming.
Intermediate
Composing agents: conversation memory, multi-agent delegation, testing, evals, graphs, and tracing.
Advanced
Production concerns: human approval, MCP as a client, model fallback, and durable execution.