Introduction

What OpenBook is, who it's for, and what makes its workflow engine different.

OpenBook is an open-source, self-hostable platform for building, running, and orchestrating AI agents and workflows. It combines a visual, drag-and-drop workflow canvas with a genuinely agentic runtime: agents call their own tools mid-run while the engine orchestrates branching, error handling, variables, and human approvals around them.

OpenBook is MIT-licensed. Run it on your own infrastructure, bring your own model keys, and extend it with plugins. There is no required cloud and no per-run pricing.

What you can build

  • Scheduled digests — fetch from RSS/Reddit/HN, score and rank with an agent, email a formatted digest.
  • Support agents — a customer message triggers a workflow that looks up an order over HTTP, branches on the situation, and routes risky actions to a human for approval.
  • Multi-agent pipelines — a chain of specialist agents that hand structured work to each other.
  • Event-driven automations — fire a workflow from a webhook or a named domain event like order.updated.

Why the engine is different

Most "AI workflow" tools can draw a graph but can't actually branch, can't recover from a failed step, and can't pause for a human. OpenBook's executor does all three — and treats agents, memory, and plugins as first-class.

CapabilityWhat it means
Conditional branchingcondition nodes route the true/false branch and prune the other — not every node fires.
Error pathsA node that throws routes to an error branch instead of failing the whole run.
Variablesset_variable writes a workflow-scoped bag; read ${vars.x} anywhere downstream.
Human-in-the-loophuman.handoff pauses the run and resumes on approval / input / choice.
Core memoryEvery agent gets long-term memory by default — no plugin to install.
AI builderDescribe a workflow in English; it creates the agents and wires the graph.

Next steps