We have an existing spike branch for a standalone assistant tab that introduces an AssistantWorkspaceTab, adds an Assistant workspace type, wires the tab into navigation and workspace opening flows, and exposes it as a first-class workspace in Compass.

Why this should be a separate task

This work addresses a different product direction than COMPASS-10544. The experiment is designed to test whether increasing visibility inside the existing workflow lifts engagement by default-opening the drawer, while the standalone tab represents a separate entrypoint strategy that does not change the current workflow in the same way.

Given that the default-open experiment is currently blocked by another experiment running on the same AI Assistant panel and audience, the standalone-tab path is a cleaner candidate for a JDI track.

Spike - https://github.com/mongodb-js/compass/compare/spike-assistant-in-tab?expand=1 

Task :-

Productize the standalone AI Assistant tab from the spike branch as a dedicated entrypoint to the assistant, with final implementation scope covering:

• Define the user-facing feature shape and expected behavior with Product.

• Assistant workspace/tab registration

• Sidebar navigation entry

• Connection selection inside the tab.

• Reuse of existing assistant chat/provider state where possible

• Drawer suppression while the standalone tab is active.