Documentation / Interfaces

The Web App: Your Control Room

The review-and-approve view over Runs, Tables and Knowledge — deliberately not a primitive of its own.

The web app is deliberately not a primitive of its own — it is the review-and-approve view over Runs, Tables and Knowledge. It renders from the same registers the agent describes: a new column that Command creates shows up in the grid immediately; nothing can drift between chat and the web app.

Principle: inspection, not primary work

For day-to-day operation, Command is the faster path. The web app exists for where visual overview and care matter:

  • Reviewing what a run would do before you approve it (scope, cost, affected rows).
  • Tracing why a cell holds its value, who classified a reply and how, which knowledge version a playbook currently uses.
  • Approving or rejecting where an approval is pending.

Feed: runs awaiting approval

The feed is not its own inbox — it is the list of open Runs waiting for your approval: an import, a live column run, a send batch. Every entry shows scope, affected rows, cost cap, and a link to the detail view. Approving or rejecting is the only action the feed asks of you — everything else (changing structure, pausing a campaign) runs through the regular view or through Command.

What the web app deliberately is not

Not a place where you have to click together every campaign by hand — the structure comes from Command or from proposed templates. Not a parallel system to Command: both interfaces read and write the same workspace data, there is no state that exists only in the web app.