Activity
Activity is the log of everything the AI did for your store on its own - and why it did it. It lives at Workspace → Activity, and it is where you go when you want to know what has been happening without you.
It has two tabs. Needs you is everything waiting for your answer, from every agent, a page at a time. History is everything the agents did, newest first and grouped by day, with each agent's latest note beside it.
What does each status mean?
| Status | What it means |
|---|---|
| Done | The action ran. |
| Needs you | The AI proposed it and is waiting for your approval. |
| Running | You approved it and it is going through. |
| Reviewed | You looked at it and did not run it. |
| Dismissed | You turned it down. |
| Failed | It could not run. The entry says why. |
What can I filter by?
By the part of the product that acted: Maestro, Email, Knowledge, Content, Segments, or Listings - or All.
What is the approval queue?
A section at the top holding what Maestro proposed from this week's numbers. Approving runs an action through your normal pipelines, so an approved email is sent the way any other email is. Items that need attention sooner are marked urgent.
How do I change what needs my approval?
On the Rules page (Workspace → Rules, or the Rules button on this page), which sets how far the AI may go for each kind of action: Ask me first means it proposes and you approve, Do it and tell me means it acts and reports back here. There is a master switch for Maestro, and the setting is per action type, so you can let it act freely on some things and not others. Pause everything on the same page stops all of it at once, and Guardrails set the limits the agents work inside, whatever the settings say. See choose your autonomy.
Why did an agent run on a day it is not scheduled?
Because something changed. Besides its weekly pass, an agent looks again when Kooga sees, in your own data, something it owns:
| What changed | What happens |
|---|---|
| A keyword you track fell three or more places below its best, on two checks in a row | The SEO agent looks at it |
| A published article got far fewer search clicks than in the two weeks before | The SEO agent looks at it |
| Your first AI visibility check finished | The GEO agent makes its first pass on real answers |
| An AI visibility check named your brand at least ten points less than the one before, on the same questions | The GEO agent looks at it |
| A keyword you track reached Google's top three, on two checks in a row | You get a card |
| Your products show up in ChatGPT's shopping cards at most half as often as in the weeks before | You get a card |
| Email sending was paused because too many recipients marked it as spam | You get a card, with what is broken |
A run is about that one thing, and its note says Why now. It reads before it can act, and it proposes what you would approve on any other day. The same thing is not raised again for a week, and an agent that ran in the last day waits until the next night.
When no agent can act on it (with the agents switched off, or when today's allowance is used), you get a card saying what moved and where to look, and nothing is started. During your trial that is always what happens, except for the GEO agent's first pass on your first check.
To hear about it by email, switch on Something changed in your account's email settings. It is off until you do, and sends one email when it happens.
Why is it empty?
Because nothing has happened yet. Actions appear as shoppers browse, ask questions and order.