This course teaches agent engineering through one primary stack: LangChain for agent/tool composition, LangGraph for explicit state and durable workflows, LangSmith for traces and evaluation, and Deep Agents for long-running agent harnesses. The gap notebooks include real framework implementations alongside offline simulations that isolate the underlying algorithm.
Python 3.11 or newer is required because current Deep Agents releases no longer support Python 3.10.
uv sync
uv run jupyter labModel-backed cells expect provider credentials such as OPENAI_API_KEY.
Tracing and remote evaluation additionally require LANGSMITH_API_KEY and
LANGSMITH_TRACING=true. Keep keys in your shell or secret manager, never in a
notebook.
The dependency ranges are based on the releases verified on July 11, 2026: LangChain 1.3.13, LangGraph 1.2.9, LangSmith 0.10.2, and Deep Agents 0.5.9. See the folder-based curriculum for the current map.