CrewAI
MIT-licensed Python framework that lets several AI agents with defined roles work on tasks together as a team.
What it does
With CrewAI you define agents in Python with a role, a goal and tools, and group them into a “crew” that splits tasks among its members. “Flows” control the process: state, events and branching. Agents use tools to reach APIs, databases or local programs. As the language model you can use OpenAI, Anthropic, Gemini, Azure, AWS Bedrock or a local model via Ollama, among others. There is also a cloud platform with a visual editor for building and running agents.
Who it suits
Development teams that want to build multi-step processes with several specialised agents, for example research, lead enrichment or support. The framework requires Python skills.
Data protection for businesses
The framework runs on your own servers; data only goes to the language model you choose. Anonymous usage statistics without prompt content can be switched off with CREWAI_DISABLE_TELEMETRY=true. For the cloud platform, the terms of use allow customer data to be used for training models as well; the vendor does not mention a way to opt out there. The vendor does not explicitly offer EU data hosting and does not publicly mention a data processing agreement. The Enterprise plan allows operation in a private VPC or on your own infrastructure, plus SSO, role-based access and redaction of personal data.
Limits
Without programming skills you will not get far with the framework. Several agents quickly consume many tokens, and their results need checking. The free cloud plan is limited to 50 executions a month.