Durable execution & workflow engines
5 systems in the durable execution & workflow engines category of the AI Agent Infrastructure Landscape, grouped by maturity tier.
Tier 3 — emerging (5)
- Cloudflare Workflows Product
Durable execution engine on Cloudflare Workers that runs multi-step workflows with automatic retries, sleeping and waiting for events. The docs include a guide to building a durable AI agent.
- Dapr Agents Framework
Python framework, GA at v1.0, for building LLM agents that run durable tasks and collaborate through Dapr Workflows, using Dapr's state management, messaging and observability.
- DBOS Framework
Postgres-backed durable workflow library that makes AI agents and applications resilient to failures, with durable queues, cron and human-in-the-loop. It integrates with Pydantic AI, LlamaIndex, the OpenAI Agents SDK and the Vercel AI SDK.
- Golem Product
Agent-native durable runtime that runs each agent in its own WebAssembly VM. After crashes, restarts or rescaling the agent resumes mid-function with memory and files intact, and tool calls execute exactly once under enforced permissions.
- Hatchet Product
Orchestration engine for AI agents, background tasks and durable workflows, available self-hosted or as a managed cloud.