Permissions
Integrated permission model for assistants, teams, and integrations
The second core topic besides the tool-call model is the integrated permission system.
Core principle
- Assistants can be shared internally with users, groups, and teams
- Integrations/connections are attached to assistants in a controlled way
- Access remains traceable and secure even when assistants are shared
Granularity by integration type
Permission control can apply on multiple levels:
- Assistant level: Who can use or edit an assistant?
- Team/group level: Which teams may use which integrations?
- Connection level: Which specific connection may be used by which assistant?
Result
This model scales from simple team setups to strict enterprise governance requirements.