The open source, multi-agent platform

CrewAI agents can now plan through long-running, complex tasks and recover automatically when errors get in the way.

CrewAI provides AI agents with hundreds of open-source tools out of the box — search the Internet, interact with websites, query vector databases, run code, and much more. With first-class support for MCP (Model Context Protocol), Custom MCP Servers, and native sandbox tools for E2B and Daytona, agents can safely execute code and call any tool you need, anywhere.

CrewAI’s memory system is built for real production agents going beyond search-backed storage, acting as a cognitive layer that intelligently remembers, resolves contradictions, forgets intentionally, and knows when it lacks context. Updated for extreme performance, simplicity, and bleeding edge workflows.

CrewAI transforms a set of AI agents into a crew of AI agents that supporting protocols such as MCP for tooling and A2A for agent collaboration and discovery

Checkpointing is a core pillar for deploying agentic systems to production. It lets users replay from specific steps, fork workflows, and run tasks with different inputs without starting from scratch.

Run crews, flows, tasks, tools, knowledge, memory, and LLM calls asynchronously with native async/await support. Streaming results flow back in real time from flows and crews, and async context managers make resource cleanup automatic. Write the orchestration you want — CrewAI handles the concurrency.

Run crews, flows, tasks, tools, knowledge, memory, and LLM calls asynchronously with native async/await support. Streaming results flow back in real time from flows and crews, and async context managers make resource cleanup automatic. Write the orchestration you want — CrewAI handles the concurrency.