How it works.
AuditMyAgent reads the records your providers already keep about AI apps and agents, turns them into a clear picture of access and activity, and gives you the controls to act on it. Here is the whole thing, including what it can't do.
From connection to control
- 1
Connect
A Google Workspace administrator signs in with Google and approves read-only access. Two minutes, nothing to install, nothing changes for employees.
- 2
Collect
We read the last seven days of Google's administrator records and then keep collecting every few minutes. Every record is stored with its original provider entry.
- 3
Review
Connected AI apps, their permissions, the people they cover, and what they did. Grouped so it reads like a log, with evidence one click away.
What you find
The dashboard opens on AI activity across the whole company and lets you narrow to one tool or one person.
Connected AI apps
Each app and agent, who connected it, when, and how many people it covers. You can mark any app as an AI tool or not, and the audit remembers.
What each app may do
Read or send mail, read or edit files, share outside the company, see calendars, administer the Workspace. Plain language first, technical name alongside.
What each app did
Files opened, downloaded and shared. Mail read and sent. Per agent, per person, per week, down to the individual record.
Who is exposed
A card for every current member of your Workspace, with their AI activity and the apps acting for them. People who have left are kept out of the default view.
Things worth a look
New authorizations, external sharing, apps granted by people who left, and access that nobody has used in weeks.
Ask in plain English
Search the history or ask a question. Every answer cites the records it came from. Weekly summaries by email, Slack or Telegram.
What you can do about it
Seeing is the Monitor plan. Acting is the Control plan. Both work from the same connection; Control adds one optional Google permission that you enable yourself.
Remove an app's access
Pick an app and a person, see exactly what will be removed, and confirm. The app loses its access for that person. Nothing else changes, and the action is recorded with who did it and when.
Set policies
Describe what an agent is allowed to do: which people it may act for, which files or folders it may touch, which actions are off limits. When new activity breaks the rule, you get a finding with the evidence. You choose the response: just record it, show the fix, or revoke the app's access for that person automatically.
Keep a record
Every control action and every policy finding is kept with its evidence, so you can show what was found, what was done, and why.
Three responses to a finding
You choose how far to go
- 1MonitorRecord the finding with its evidence. Nothing changes.
- 2GuideRecord it and show the exact fix in Google Admin.
- 3AutomaticRevoke the app's access for that person as soon as the activity is reported. Google Workspace only, and only after you confirm the rule.
Good to know
- !Revoking access doesn't stop someone reconnecting the app later. To block an app for good, use the app access setting in Google Admin. The dashboard points you to it.
We show you everything your provider recorded.
We don't pretend to see what it didn't.
The honest boundaries
It works from provider records. Google writes down app connections and Drive, Gmail and Calendar activity in its administrator logs. That is what we read. We never open an email or a file, and we never see keystrokes or screens.
Logs can lag. Google publishes some records within minutes and others a few hours later. The dashboard shows how current each source is, so you always know what you're looking at.
Attribution is only as good as the record. When Google names the app behind an action, so do we. When an app acts on a person's behalf, the record often shows the person. We label the difference instead of guessing.
Monitoring is after the fact. A policy can revoke access the moment an action is reported, but it cannot un-send an email or recall a file that already left. For agents you run yourself, AgentGuard also offers a separate approval gateway that holds risky actions until a human says yes. Ask us about it.
Google Workspace today. Microsoft 365, Slack and GitHub are built and in testing with early customers. Each will show the same kind of inventory, permissions and activity, within what that provider records.
See it on
your own Workspace.
Connect as an administrator and have your first findings within the hour.
Audit My WorkspacePricing starts at $5 per employee per month, with a 3-day free trial of Control.