CrewAI
Open-source Python AI agent framework under MIT. Run workflows on your own infrastructure with Ollama models or connect to the OpenAI API.
CrewAI is an open-source Python framework for developers building workflows with multiple AI agents on their own hardware or servers. It supports local models through Ollama and defaults to the OpenAI API for model requests. The framework uses the MIT license; a separate commercial platform provides managed deployment and governance, with on-premise and cloud options.
Its two main building blocks give you different levels of control. Crews group specialized agents with roles, goals, tools and assigned tasks. Flows put agent work inside event-driven workflows with state, branching and explicit business logic. You can combine both in one application, so autonomous decisions can sit alongside predictable steps and human review.
Agents can use memory and knowledge, return structured results, and save checkpoints. MCP and A2A support let them connect to tools and other agents. Python APIs give developers control over prompts and execution logic, while the commercial platform also has a visual builder that exports to Python.
For teams managing agents in production, the commercial control plane traces model and tool calls, tracks costs, and adds access controls, audit trails and approval gates. Runtime checks can redact personal data and enforce policies.
Model choice affects where requests go: Ollama supports local inference, while the OpenAI API sends model requests to OpenAI. The framework also collects anonymous usage telemetry, which you can disable. Enabling share_crew allows it to collect task descriptions and agent attributes.
