Multi-agent framework

OpenBook vs AutoGen

AutoGen is Microsoft's code-first framework for multi-agent conversations and tool use, with strong primitives for agent-to-agent collaboration. OpenBook is a self-hostable platform: a visual + agentic canvas, an AI builder, persistence, a plugin ecosystem, long-term memory, and human-in-the-loop — without writing orchestration code. AutoGen suits Python teams embedding multi-agent logic; OpenBook suits teams who want a runnable product.

At a glance

OpenBook and AutoGen, capability by capability.

CapabilityOpenBookAutoGen
Open source
Visual builder + AI builder
Runnable product (UI, persistence, runs)
Multi-agent orchestration
Long-term memory by defaultConfigurable
Human-in-the-loop pause/resumeHuman input (code)
Plugin / action ecosystemTools via code
Embed as a Python library

Why teams pick OpenBook over AutoGen

A platform, not a library

OpenBook ships a UI, persistence, run history, an approvals inbox, and a plugin system. AutoGen is code you wire into your own app.

Visual + AI builder

Build on a canvas or let the AI builder generate the graph — no Python required. AutoGen multi-agent setups are defined in code.

Human-in-the-loop built in

Pause a run for approval and resume from an inbox. AutoGen supports human input, but you build the surrounding product yourself.

Memory + shared actions pool

Agents remember by default and draw from one shared pool of Actions and Skills, extended by plugins.

OpenBook

MIT platform you self-host.

AutoGen

MIT framework (Microsoft) you embed in code.

When to choose which

Choose OpenBook when…
  • You want a runnable, operable product — UI, persistence, approvals — not just a library
  • You want non-engineers to build with a canvas or the AI builder
  • You need built-in human approvals and memory
Choose AutoGen when…
  • You're a Python team embedding multi-agent conversations in your own application
  • You want code-level control over agent-to-agent collaboration patterns
Switching from AutoGen

Map AutoGen agents to OpenBook agents (system prompt + actions) and conversation patterns to a workflow graph of agent nodes. Human-input steps become human.handoff nodes.

Frequently asked questions

Build your first agentic workflow today.

OpenBook is free and open source. Self-host it in minutes, or read the docs to see how far the engine goes.