Example 13: Durable Workflow Agent

This example shows checkpoint and resume for long-running agent workflows.

Run

python examples/13-durable-workflow-agent/main.py

What It Shows

• Save workflow state after every step
• Stop partway through a run
• Resume from the latest checkpoint
• Preserve artifacts across a restart
• Verify that the resumed run completed safely

Learning Check

• Which state must be saved to resume safely?
• Which tool calls are safe to retry?
• Which side effects need idempotency keys?