OpenAI Agents SDK
All frameworks
Vendor SDKs · by OpenAI

OpenAI Agents SDK

OpenAI's own agent loop, with handoffs, guardrails and tracing built in.

27 lessons2h 29m reading8 sections
01Your first agent0/7 · 33m
OpenAI Agents SDK overview4m
Setup: install the openai-agents package3m
Agent: name, instructions and tools4m
Model: a stand-in you can run without a key6m
Runner and the RunResult object5m
function_tool: a function the model can call5m
How the agent runs a tool then answers6m
Section quiz · 4 questionsNot taken
02Answers your code can trust0/3 · 17m
Structured output with output_type6m
Passing data with RunContextWrapper6m
Instructions from a function5m
Section quiz · 3 questionsNot taken
03More than one agent0/3 · 17m
Handoffs: passing control to another agent6m
An agent as a tool with as_tool6m
Handoff vs agent-as-tool5m
Section quiz · 3 questionsNot taken
04Keeping it safe0/2 · 12m
Input guardrails and the tripwire6m
Output guardrails on the final answer6m
Section quiz · 3 questionsNot taken
05Memory0/1 · 6m
Sessions: memory across runs with SQLiteSession6m
Section quiz · 2 questionsNot taken
06Watching and shaping a run0/4 · 23m
Streaming a run with run_streamed6m
Tracing a run with spans6m
Shaping the model with ModelSettings5m
Capping a run with max_turns6m
Section quiz · 3 questionsNot taken
07The project: a shop support desk0/5 · 32m
The desk agent and its tools6m
A refund specialist by handoff6m
Guarding the desk's input and output6m
The desk that remembers with a session6m
A full run, and when it refuses8m
Section quiz · 3 questionsNot taken
08Off the stand-in0/2 · 9m
Swapping in a real model5m
What we left out, and what is next4m
Section quiz · 2 questionsNot taken