Activation controls whether an Area’s instructions and attachments are part of the current Agent context. It does not change the underlying permissions on the Sources, Standards, or Connectors inside it.
Always on
Core scope for every run.
- Starts active
- Cannot be deactivated
- Best for identity and universal duties
- Consumes context on every question
On demand
Specialized scope loaded when relevant.
- Starts inactive
- Can be selected by the person or planner
- Best for distinct domains
- Leaves unrelated context out
Choose always on sparingly
Use it for behavior or context the Agent needs for nearly every request. Making every Area always on removes the benefit of composition and can make instructions compete.
Describe on-demand triggers
Write a description that names the questions, tasks, or conditions that should activate the Area. Only an Area that has a description appears in the routing catalog at all; one without a description is invisible to routing.
What activation changes
The Area’s instructions join the Agent’s direct instructions, and its Knowledge, Standards, and Connectors join the runtime scope. Deactivation removes that contribution; direct Agent attachments remain.
What activation cannot override
A reader still cannot use a Role-restricted Standard they may not read. A tool the organization marked as requiring confirmation still goes through a proposal, and one the platform reserves for people is never offered at all. Activation composes available material; it does not widen authorization.