Assistant capabilities

Capabilities are per-assistant permissions and configuration blocks.

They help define what an assistant is allowed to do and, in some cases, which actions it may try to run.

What a capability is

In the app, a capability can serve one or more of these roles:

it can appear as assistant configuration metadata;
it can gate a behavior in the prompt or runtime; or
it can map to an executable handler or bridge.

Not every capability in the registry should be read as a fully live customer-facing action.

Scope

Capabilities are scoped to one of these categories:

organization for Chiro-only capabilities;
user for personal EA capabilities; or
both for capabilities that can apply to either type.

What is live in the app

The app includes:

a capability registry;
assistant-specific enabled capability lists;
tier-based filtering and limits; and
assistant settings screens for managing capabilities.

Some capability IDs also have executable runtime handling. Others should be treated as configuration or registry metadata unless the product clearly exposes the action as live.

How to document capabilities safely

When writing docs, treat capabilities as product controls first, not as proof that every listed action is fully available in every workspace.

Use these companion pages for the practical product view: