Yes this is a microservices arch with a built in acp where the permissions define what each agent/user can see. Then you can combine inference + tools + environment on the fly for each need. You’d prob want it cryptographically verifiable and secure by design. I’ve been working on this.