Building CrewAI agents: How to secure Crew AI agents?

We are working on building CrewAI agents and wondering how to secure AI agents and workflows.

  • How do we go about managing API keys and authentication?
  • How do we handle identity and credential management?

Any recommendation or opensource tools that can help build our Crews secure by default will be helpful.

Hi @imaxxs and welcome to the community.

Good idea to store your keys in .env and for secure deployment you can use crewAI or as it is python secure it as you would you pythons environment as there are lots of patterns for this.

For MCP this is worth a read MCP Security Considerations - CrewAI

1 Like

Thank you, Tony. Storing keys in .env file for prototyping may be ok but not for production readiness. What do you mean by - “for secure deployment you can use crewAI”? How does crewAI help with secure by default or authentication and authorization of the resources agent needs to access?