Catalog & capabilities
Browse the discovered capabilities and refine each one — its description, permissions, risk, and more.
The catalog is the list of everything the assistant can do, discovered from your API. This is where you review it and tune each capability.
Browsing the catalog
Capabilities are grouped by domain, each with its HTTP method, side effect, and an internal badge where relevant. Search by name, and filter by method, by side effect, or to hide internal capabilities.
Enable and tune a capability, step by step
A capability has to be reviewed and made active before the assistant will use it. Here is the whole loop:
- Open the capabilityFind it in the catalog — search by name or filter by method — and open it to reveal its editor tabs.
- Give it a clear descriptionOn the General tab, set a plain display name and a description of what it does. A good 'when to use this' hint on the Prompt tab helps the assistant pick it at the right moment.
- Set its riskOn the Risk tab, confirm or override the side-effect class (Read, Write, or Destructive) and set a risk level (Low to Critical). This decides whether the action runs automatically, asks for confirmation, or stays blocked. Edits here save automatically.
- Enable it on the Governance pageNewly discovered capabilities start disabled. Open the Governance page and switch the capability to Enabled so the assistant can use it — and optionally turn on Require approval. Until it’s enabled, the assistant can’t see or call it.
Editing a capability
Open a capability to edit it across a few tabs:
- General — the display name, description, and whether the capability is internal or external.
- Permissions — the discovered scopes a user needs, plus a map to rename scopes.
- Schema — the request and response shapes, shown read-only.
- Prompt — a 'when to use this' hint that guides the assistant.
- Risk — override the side effect and set a risk level, from low to critical.
- Audit — the full change history, with a diff and one-click revert.
Auto-save & history
Your edits save automatically as you make them. If someone else changes the same capability, a banner lets you reload rather than overwrite their work — and the Audit tab keeps the full history so you can always look back or revert.