CrewAI

Active
GitHub Python MIT

Description

CrewAI is a multi-agent framework for orchestrating role-playing, autonomous AI agents that collaborate like a team to tackle complex tasks.

Key Features

  • Role-playing agents — every agent has role, goal, backstory and a private toolset
  • Task orchestration — express dependencies, parallelism and context handoff between tasks
  • Autonomous collaboration loops — delegation, hierarchical and consensual processes built-in
  • Tool ecosystem — 100+ pre-built tools plus LangChain, Composio and MCP server integrations
  • Memory & knowledge — short-term, long-term, entity memory and knowledge base retrieval
  • Training & evaluation — crewAI train command and evaluation harness for feedback-driven improvement

Use Cases

💡 Building market research, code review, content production multi-agent pipelines
💡 Decomposing complex business work into sub-tasks with auto-delegation and aggregation
💡 Standing up a research team with role division on top of an enterprise knowledge base
💡 Combining with LangChain / LlamaIndex for tool-using multi-agent workflows
💡 Continuously improving agent output quality via feedback-driven training

Strengths & Limitations

Strengths

  • Actively maintained, recent updates
  • High community interest (58.0k stars)
  • Permissive open-source license (MIT)
  • Established track record (2 years in production)

Quick Start

# Install the SDK
pip install crewai crewai-tools

# Define a researcher and a writer agent
from crewai import Agent, Task, Crew
researcher = Agent(role='Researcher', goal='Gather latest AI industry news',
                  backstory='Senior industry analyst', tools=[search_tool])
writer = Agent(role='Writer', goal='Turn research into a short article',
               backstory='Tech columnist')

# Describe the tasks
research_task = Task(description='Survey 2026 Agent framework trends', agent=researcher)
write_task = Task(description='Write a 500-word article from the research', agent=writer)

# Kick off the crew
crew = Crew(agents=[researcher, writer], tasks=[research_task, write_task])
result = crew.kickoff()

Related Projects

Hive

11.0k · Python
Active B

Hive is a production-ready multi-agent execution harness providing state management, failure recovery, observability, and human-in-the-loop control with auto-generated multi-agent topologies for complex business workflows.

multi-agentagent-frameworkproduction +3
  • · Zero-setup, model-agnostic execution harness that auto-generates multi-agent topologies for complex workflows
  • · Graph-based execution DAG for safe parallel task coordination with deterministic fault tolerance
  • · Persistent role-based memory that evolves with project context — no manual state management

AgentScope

30.5k · Python
Active A+

Open-source multi-agent framework from Alibaba, enabling the construction of observable and interpretable agents with rich distributed capabilities.

agent-frameworkmulti-agentdistributed +2
  • · Built-in ReAct Agent — Create agents with reasoning and tool-use capabilities in 5 minutes
  • · Tool & Skill System — Built-in Bash, Grep, file I/O tools with custom extension support
  • · Multi-Agent Orchestration — Flexible multi-agent collaboration via Message Hub

OxyGent

2.1k · Python
Normal B

OxyGent is an open-source multi-agent collaboration framework from JD.com, supporting flexible agent role definition, task decomposition, and collaborative orchestration for enterprise AI agent applications.

multi-agentcollaborationenterprise +2
  • · Modular multi-agent framework with LEGO-brick-style Oxy components for rapid assembly
  • · Dynamic planning with intelligent task decomposition, negotiation, and real-time adaptation
  • · Elastic architecture supporting any agent topology from simple ReAct to complex hybrid planning

Langroid

4.1k · Python
Active A+

Langroid is a Python multi-agent programming framework that leverages an intuitive Agent-Task-Tool abstraction to help developers build LLM-powered multi-agent applications.

multi-agentllmagent-framework +1
  • · Multi-agent programming: intuitive Agent-Task-Tool abstraction inspired by the Actor Framework
  • · LLM-agnostic: works with practically any LLM, including local models via OpenAI-compatible APIs
  • · MCP tool adapter: convert MCP server tools into Langroid ToolMessage instances

Related Articles