Durable Agent Workflow
A persisted, code-governed graph of steps, state, retries, approvals and terminal outcomes for an AI workflow that must survive interruption or be audited.
A durable agent workflow makes execution explicit. Nodes represent permitted steps or states; edges represent allowed transitions. The run retains checkpoints, retry rules, cancellation, idempotency and a record of approvals.
Use it for work that survives restarts, waits for a human decision, has parallel branches, triggers side effects or needs an audit trail. A graph is the backbone; it does not remove the need for bounded agent reasoning and policy checks.
Still curious?
Ask Kuni, your AI learning companion, to explain this concept in the context of your own work.
AI can make mistakes. Check important facts, decisions and sources before relying on them.
Ask MAESTRO
Ask FORGE
Ask ORION
Ask SENTINELSTART
Bounded roles
CONTROL
Visible hand-offs
OUTCOME
Human owner
AGENTS & WORKFLOWS
Go deeper
Related concepts
Seen in cases
Real-world examples where this concept appears in our case studies.
Ready to bring AI into your organization?
Talk to us about a guided adoption path for your team — from first use case to production.
Ask about this concept