- Who deleted that table, or rotated that secret?
- What changed just before errors started?
- Which key or AI tool made a change, and when?
- Did someone restore a backup or release an API version?
What each entry contains
Values are never recorded. The log says that a secret changed, not what it changed to. Arguments an AI tool passes are recorded by name only.
What gets recorded
The table lists the most common actions; other configuration changes are recorded the same way.
Use it
After an incident
- Note the time errors started in Activity.
- Open the audit log and look at what changed in the minutes before.
- If a change is the cause, undo it, or restore from a backup.
Find who did something
Read down the Action and Target columns, newest first. Because each integration should have its own key, the actor tells you which app or job made a change, not only which person.Review AI tool activity
Everything an AI tool does shows up here asmcp.tool and on the AI → AI tools page. Review after you let a tool build with a Build with me key.
How long it is kept
Audit history is kept for a time that depends on your plan:
If you need to keep it longer, copy what matters elsewhere on a schedule.
Good practice
- One key per integration, so the actor means something.
- Don’t share a Studio login. Invite each person to the organization (Team).
- Look at it after every release, alongside errors.
- Pair it with alarms for the changes that must never happen unnoticed.
What it does not show
- Data changes made through your app’s API (a customer editing their own record). Those are your application’s data, not configuration. Use Requests, or an event subscription if you want your own history of record changes.
- The contents of anything secret.
- People signing in. See Sign-in activity.