CrewAI is an open-source Python framework for orchestrating role-playing, autonomous AI agents through collaborative Crews and event-driven Flows, released under the MIT licence and built for developers and teams producing multi-agent workflows in production.

What it is

CrewAI is a Python framework that combines high-level abstractions with low-level APIs for assembling production-ready multi-agent workflows. It is organised around two primitives. Crews optimise for autonomy and collaborative intelligence using role-based AI agents, each configured with a role, goal, backstory, tools, LLMs, memory, and guardrails. Flows provide event-driven automation that mixes precise workflow control, single LLM calls, and native support for Crews, so the same project can move between emergent agent collaboration and deterministic control. The repository carries 58,726 stars, 8,498 forks, and 353 open issues, with the most recent push on 18 September 2026, and it is tagged across the agent and LLM topic space as agents, ai, ai-agents, aiagentframework, and llms.

The concrete problem it addresses is the glue code that otherwise accumulates when a team wants several language models to work on one task. Rather than hand-wiring prompts, sequential single LLM calls, retries, and custom event plumbing, a developer declares agents and tasks and lets a Crew handle delegation, or declares a Flow when the order of operations must be explicit. It lives in the Python LLM and agent ecosystem, and it replaces the bespoke multi-agent orchestration layer that teams would otherwise write and maintain themselves.

Key capabilities

• Crews for autonomous, role-based agent collaboration, with per-agent configuration of role, goal, backstory, tools, LLMs, memory, and guardrails.
• Flows for event-driven automation that combines precise workflow control, single LLM calls, and native support for Crews in one program.
• Structured task output through output_pydantic and output_json, with support for task dependencies and human review.
• Project scaffolding wired through crew.jsonc and main.py, covering the README's Installation, Setting Up Your Crew, and Running Your Crew steps.
• Claude Code integration via /plugin marketplace add crewAIInc/skills, /plugin install crewai-skills@crewai-plugins, and /reload-plugins, activating the getting-started, design-agent, design-task, and ask-docs skills.
• Cross-editor skill installation for Cursor, Codex, Windsurf and others with npx skills add crewaiinc/skills, plus a live docs MCP server at https://docs.crewai.com/mcp.
• Worked examples in the README, including Write Job Descriptions, Trip Planner, and Stock Analysis, plus a documented telemetry section.

Who uses it and how

• Python developers building production-ready multi-agent workflows who need role-based delegation and explicit workflow control in the same codebase.
• Enterprises that layer the CrewAI AMP Suite around the open-source framework for managed deployment, observability, governance, security, and enterprise support.
• Platform and operations teams that need tracing and observability over agents and workflows, including real-time metrics, logs, and traces.
• Learners and community members: the project reports over 100,000 developers certified through courses at learn.crewai.com.
• Teams already working inside an AI coding agent, which teaches CrewAI conventions through the official CrewAI Skills package.

Getting started

The README's Installation section covers installing the Python framework, followed by Setting Up Your Crew and Running Your Crew to scaffold and execute a first project. A hosted option is also available: the Crew Control Plane can be tried free at app.crewai.com.

How it compares

No comparable open-source orchestration frameworks are named in the facts, so on this registry the project stands alone rather than being contrasted against named alternatives. The nearest related offering is the vendor's own CrewAI AMP Suite, which is a commercial control plane rather than a replacement for the MIT-licensed framework: it adds managed deployment, on-premise or cloud hosting, advanced security and compliance measures, and 24/7 support, while the framework itself remains free and self-hostable.

When to use it — and when not to

Choose it when the team writes Python, wants MIT-licensed code it can run itself, and is prepared to connect its own model access and operate its own runtime, since a crew must be connected to a model before it does anything. Teams that need governance, compliance controls, or guaranteed support should weigh the commercial AMP Suite rather than treating the open-source package as complete on its own. Be aware that the surfaced README is largely a table of contents and suite description: the installation command, telemetry behaviour, and licence text sit behind sections not reproduced here, and 353 open issues indicate an active but busy tracker — check current docs and release notes before committing a production system to it.

/plugin marketplace add crewAIInc/skills
/plugin install crewai-skills@crewai-plugins
/reload-plugins

npx skills add crewaiinc/skills