Zinits Agent is the governed AI surface inside your workspace.

It combines a selected Agent, an available model source, workspace permissions, approved tools, and durable context. It is not a general promise of autonomous action.

Model access

A workspace may use Zinits-included access, a sponsored source, or a provider connection supplied by the workspace. The current selection and readiness state should be visible. Included access during a controlled pilot does not promise unlimited use or continuous availability.

Provider choices

Connected-provider choices can vary by workspace. Credentials are handled by the authenticated application and are not published to the browser as plain configuration. Switching sources does not silently delete another connection.

Memory and context

Agent conversations can use permitted workspace context and durable goals. Context is bounded by the active tenant, Agent, grants, and tool policy. A previous conversation does not override current permissions.

Approvals and effects

Drafting and analysis do not authorize delivery, submission, deletion, access changes, or deployment. External effects pause at the applicable confirmation or approval boundary.