Auth and Permissions Delegation Layer for Crew AI Agents

[imaxxs]

Hi,

I reviewed some of the security features currently available for the CrewAI builders and users.
I started with reading the code here: https://github.com/crewAIInc/crewAI/blob/main/src/crewai/security/security_config.py

I would love to see a lot of secure by default capabilities around auth, user-agent delegation, agent-agent delegation, policy enforcement, and more be added to CrewAI agents and AI workflows.

I am curious if there is an interest to work together and add many of these feature implemented as a open source module.

I’m building DeepTrail (Berkeley SkyDeck-backed). At Deeptrail, we are building open source auth and delegation layer for AI agents. With Deepsecure (Open source), developers effortlessly integrate authorization, agent-to-agent delegation, policy enforcement, and secure proxying - across any model, platform, or framework - with just a few lines of code.

Technology: Split-key architecture, gateway/proxy, separate control/data plane, policy engine, and macaroons for agent-agent delegation.

Deepsecure open source: https://github.com/DeepTrail/deepsecure.
Technical overview: https://github.com/DeepTrail/deepsecure/blob/dev/docs/design/deepsecure-technical-overview.md

I have also built some simple examples and integrations for CrewAI:

1. Delegation Workflow - https://github.com/DeepTrail/deepsecure/blob/dev/examples/10_crewai_delegation_workflow.py
2. Advanced Delegation Patterns -
   https://github.com/DeepTrail/deepsecure/blob/dev/examples/11_advanced_delegation_patterns.py
3. Secure Multi-agent workflows with fine grained access controls -
   https://github.com/DeepTrail/deepsecure/blob/dev/examples/03_crewai_secure_tools.py

I would love to see if this is something founders and core contributors would be interesting in discussing and exploring together.

Mahendra Kutare
https://github.com/imaxxs/
https://www.linkedin.com/in/imaxxs/