What is durable workflow execution?

Definition

Durable workflow execution records process state outside a single running program. If a worker stops, a service restarts, or an external call fails, the workflow can resume from a known point instead of losing all progress.

This is important for agent workflows that coordinate research, drafting, scoring, review, and human tasks. Durable state supports retries, timeouts, checkpoints, and isolation of individual failures.

Acronyms and aliases

durable workflow variant

Frequently asked questions

Why do artificial intelligence agents need durable workflows?

Agent work often spans several services and review steps, so durable state prevents one interruption from destroying the entire process.

Is durable workflow execution the same as a task queue?

No. A queue distributes work, while durable execution also tracks sequence, state, retries, waits, dependencies, and outcomes.

Videos explaining durable workflow execution