Guide: Permissions

Practical permissions strategy for integrations in chats and assistants

Goal

Start securely without blocking delivery through overcomplicated permission models.

Practical rules

  • begin with a few assistants and clear audiences
  • attach integrations only where they are truly needed
  • separate teams/groups clearly (sales, support, ops)

Connector or data pool?

The conceptual permission model — which level (assistant, team, connection) decides what — is described under Permissions. For day-to-day work, this rule of thumb is enough:

  • operational live workflows -> connector-first (carries over user-level permissions from the source system)
  • large-scale knowledge retrieval -> data pool with explicit access controls (centrally managed access, native user permissions do not carry over automatically)
Was this page helpful?