# OpenProgram vs LangGraph, AutoGen, and CrewAI

This page compares the public programming models documented by each project. It
is intended to help developers choose an abstraction, not to rank unrelated
features. The comparison was last verified on **2026-08-13** against the linked
official documentation.

| Framework | Documented primary abstraction | Documented orchestration model | Documented emphasis |
|---|---|---|---|
| **OpenProgram** | `Agent` method plus an execution runtime | Ordinary control flow, model-selected tools, and a shared execution DAG | Agents that can author reviewable agentic functions; runtime-managed context, tools, memory, interfaces, and multi-agent work |
| **LangGraph** | Graph API with nodes and edges, or Functional API with `@entrypoint` and `@task` | Explicit state graphs or standard Python control flow over the same runtime | Durable execution, persistence, streaming, and human-in-the-loop control |
| **AutoGen** | AgentChat agents and teams, or Core agents and runtimes | Agent messages, team patterns, and an event-driven Core API | Conversational single/multi-agent applications and scalable multi-agent runtimes |
| **CrewAI** | Agents, tasks, crews, and flows | Role-based agent crews combined with event-driven flows | Collaborative agent teams plus structured workflow automation |

Sources: [OpenProgram Agentic Programming](../capabilities/agentic-programming/README.md),
[LangGraph overview](https://docs.langchain.com/oss/python/langgraph/overview),
[LangGraph Functional API](https://docs.langchain.com/oss/python/langgraph/functional-api),
[AutoGen overview](https://microsoft.github.io/autogen/), and
[CrewAI documentation](https://docs.crewai.com/).

## Choose OpenProgram when

- an agent should be able to propose or author a new workflow as a reviewable
  function rather than only select from a fixed graph or team configuration;
- one runtime should provide terminal, Web, model-provider, tool, memory,
  context, and multi-agent surfaces;
- execution context should be represented as a DAG of user, model, function,
  and tool calls.

Start with [Self-Programming AI Agents](../capabilities/agentic-programming/self-programming-ai-agents.md) and
the [OpenProgram installation guide](../start/GETTING_STARTED.md).

## Choose LangGraph when

Your system needs durable execution, persistence, streaming, and human-in-the-loop
state control. LangGraph provides an explicit Graph API and a Functional API that
supports ordinary Python branches, loops, and function calls. Its overview
describes LangGraph as a low-level orchestration runtime and recommends
higher-level LangChain agents for prebuilt agent architectures.

Official sources: [LangGraph overview](https://docs.langchain.com/oss/python/langgraph/overview)
and [Functional API](https://docs.langchain.com/oss/python/langgraph/functional-api).

## Choose AutoGen when

Your design is centered on agents exchanging typed messages or participating in
team patterns. AutoGen exposes a higher-level AgentChat API and a lower-level,
event-driven Core runtime; its documentation covers teams, state management,
human feedback, custom agents, and distributed runtimes.

Official sources: [AutoGen](https://microsoft.github.io/autogen/) and
[Agent and Agent Runtime](https://microsoft.github.io/autogen/stable/user-guide/core-user-guide/framework/agent-and-agent-runtime.html).

## Choose CrewAI when

Your application maps clearly to agents with roles, goals, and tasks organized
as a crew, or to structured event-driven flows that invoke crews for autonomous
work. CrewAI documents crews for collaboration and flows for controlled workflow
execution.

Official source: [CrewAI documentation](https://docs.crewai.com/).

## Verification boundary

The table summarizes documented public abstractions; it does not claim that an
unlisted feature is impossible through extensions or custom code. Project APIs
change, so verify the linked documentation before making a long-term migration
decision.