Execution activity
This view is a ledger of every agent run across the workspace. Expanding a run shows its full timeline: the plan the agent formed, each tool call it made, and any verification steps, along with which surface it came from: the Assistant chat, the floating chat launcher available on other pages, a document, an automation, WhatsApp, or Canvas.
Capability registry
This view lists every tool the AI is able to call, grouped by domain, each tagged by risk: Read, Write, or Always review. It's the reference for exactly what the agent can and can't do, and which of those actions are read-only versus consequential.
- Execution activity: expandable per-run plan / tool-calls / verification timeline
- Capability registry: every tool, grouped by domain, tagged Read / Write / Always review
What this module is not
AI Tools has no approve, reject, edit, or run action anywhere in it. It exists purely for visibility. If you're looking to act on a pending action, that's the Approvals inbox; AI Tools links directly out to Settings → AI usage and to the Assistant for when you do need to act elsewhere.
How to
How to trace what an automation or chat actually did
- Open AI Tools and switch to Execution activity.
- Find the run by its surface (assistant, automation, WhatsApp, canvas, etc.) or by searching.
- Expand it to see the full plan, every tool call it made, and any verification recorded against it.
Good to know
- This module is purely observational. Treat it as the audit trail. Any decision about whether an action should happen is made in Approvals, not here.