Skip to main content
The Policies tab controls what Logic AI can and cannot do in your org. This is where you define tool permissions, per-call limits, approval requirements, and which models users can pick.

​
How Policies Work

Logic AI uses a layered policy system:
  1. Org default — Applies to every chat user
  2. Per-user overrides — Take precedence over the org default for individual users
When a user starts a chat, Logic AI resolves their effective policy by checking for a user-specific override first, then falling back to the org default.

​
Tool Permissions

Tools are grouped into Reading data, DML, and Utilities. Each tool can be individually toggled on or off: Disabling a tool removes it from Logic AI’s available actions entirely — it won’t suggest or attempt to use it.

​
Metadata (Admins only)

The List metadata and Read metadata tools appear in their own Metadata (Admins only) group. Even when enabled, they run only for users with the Logic AI Admin permission set (and reading Apex additionally requires the user’s own “Author Apex” permission). They let Logic AI read how your org is built — see the Metadata Tools reference.

​
Per-Call Limits

Some tools expose a Max per call limit:
  • SOQL query — maximum rows returned (default 50, up to 2,000)
  • Mass create / update / delete records — maximum records per operation (default 200, up to 10,000)

​
Approval Requirements

Write tools can require the user to approve the action in-chat before it runs:
  • Create record / Update record / Mass create / Mass update — approval is configurable (toggle “Needs approval”)
  • Send email — approval is required by default (configurable); Logic AI shows the recipients, subject, and body before sending
  • Delete record / Mass delete records — approval is always required and cannot be turned off

​
Model Allowlist

The Models section controls which Anthropic models the chat model picker exposes. Toggle individual models on or off. Leaving all models on means no restriction (users see the full picker).

​
Chat & Coworkers

Controls the default assistant chat and who can create coworkers. Both settings exist on the org default and per user; admins always bypass them.
  • Coworkers only — hides the default New chat so the user works exclusively through coworkers. Use it to keep everyone in structured, purpose-built personas.
  • Disable creating coworkers — hides Add coworker. Users can still use coworkers shared with them, but can’t create their own.

​
Per-User Overrides

Below the org default, every user assigned the Logic AI User permission set is listed (admins are excluded). Click a user to open their override editor, where you can:
  • Enable or disable individual tools just for that user
  • Adjust their per-call limits and approval requirements
  • Restrict the models they can pick
  • Set their chat access (Coworkers only / Disable creating coworkers)
Each user shows whether they inherit the org default or are customized. Use Revert to default to clear a single user’s override, or Revert all to default to clear every override at once.