CrewAI
A multi-agent orchestration framework where you define specialized agents, give them tools, and watch them collaborate on a task.
Think of it like
A newsroom where each reporter has a beat—one covers politics, one covers tech—and an editor coordinates the story.
Example
Three agents: researcher (searches the web), writer (drafts prose), editor (fact-checks and polishes). Tell CrewAI to 'write a piece on AI regulation in 2024,' each agent does their part, hands off to the next.
How it actually works
CrewAI abstracts agent design into roles, tools, memory, and handoff logic. You define each agent's goal and tools, it manages communication between them. Behind the scenes, it's still LLM calls and tool use, but the framework handles task routing and context passing. Feels closer to real teams than single-agent loops.
For product teams
Enables parallel skill sets and role-based automation; speeds up complex multi-step workflows.
For engineers
Graph-based task routing with role definitions; LLM-agnostic but optimized for Claude/GPT.
Read anything AI without the jargon
Look up any term in plain English, or save terms as you read with the free Chrome extension.
Open DecoderAdd to Chrome