Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/temporalio/edu-ai-workshop-openai-agents-sdk/llms.txt

Use this file to discover all available pages before exploring further.

Build durable AI agents that survive failures, retry automatically, and scale to production — powered by OpenAI Agents SDK and Temporal. This workshop guides you from your first agent to a multi-agent routing system, all running in GitHub Codespaces with zero local setup.

Quick Start

Launch your Codespace and run your first agent in minutes

Environment Setup

Configure your API keys and start the Temporal server

Core Concepts

Understand agents, tools, workflows, and durability

Exercises

Four hands-on exercises from hello world to production routing

What You’ll Build

Over four progressive exercises, you’ll go from a simple agent with a weather tool to a fully durable, multi-language routing system backed by Temporal:
1

Agent Hello World

Create your first OpenAI agent with a real weather tool that calls the National Weather Service API. Learn how LLMs decide when to invoke tools.
2

Temporal Hello World

Write your first Temporal workflow and activity. Watch execution in the Temporal UI and experience automatic retries.
3

Durable Agent

Combine the OpenAI Agents SDK with Temporal to wrap LLM calls in activities — getting retries, crash recovery, and full observability for free.
4

Agent Routing

Build a multi-agent triage system that routes queries to French, Spanish, or English specialist agents using the handoff pattern, structured as a production-ready Temporal application.

Why Temporal for AI Agents?

Automatic Retries

LLM API calls wrapped in Temporal activities retry automatically on failure with configurable backoff policies

Crash Recovery

Workflows persist their state — if your worker crashes mid-execution, Temporal resumes exactly where it left off

Full Observability

Every activity, handoff, and LLM call appears in the Temporal Web UI with complete execution history

Zero Agent Changes

Your OpenAI agent code stays identical — Temporal wraps it with production superpowers via a plugin

Workshop Structure

This is a 90-minute workshop: 30 minutes of instruction followed by four 15-minute hands-on exercises. During the workshop — work through the complete, running implementations in solutions/. These are fully working examples you’ll execute, observe, and learn from. After the workshop — practice building everything from scratch using the starter code in exercises/ as optional homework.
All exercises run in GitHub Codespaces. You need a GitHub account and an OpenAI API key (~0.500.50–1.00 of credits for the full workshop).

Build docs developers (and LLMs) love