Skip to content
Ironflow is pre-1.0: APIs can change between releases. Pin a version. What is ready →

What can you build with Ironflow?

Start with the work your application needs to finish. Ironflow is a durable runtime for AI agents and event-driven systems. It preserves completed step results, waits for external input, and supports state derived from events.

The use cases below connect those capabilities to concrete applications. Each explains a job, its difficult moment, an illustrative implementation flow, and the pieces your application supplies. The starting guides demonstrate specific runtime behaviors; some use cases combine them into an application you would build.

  • Agents over A2A: expose a specialist agent so other agents can discover it, delegate tasks, and follow progress.
  • Agent harnesses with MCP: expose application capabilities to AI clients and connect them to durable work.
  • Document processing: extract, classify, and publish documents through worker interruptions, with an optional review step you add.
  • Support resolution agents: investigate a case, obtain approval, and wait for a customer reply.
  • Reconciliation: match routine transactions and resolve exceptions through agent triage, approval, and outside responses.

An application is a useful fit when completed work must survive an interruption, the next input might arrive tomorrow, or several services must converge on one visible result. Look for the use case with the closest failure or wait to yours. Ordinary reads and writes can continue to use your application database.

For a first runnable introduction, follow Getting started. For an agent, Survive a crash demonstrates recovery after a worker is killed during a document-processing task.