Skip to main content
Every operation an MCP key runs is recorded. The log is metadata only — what was attempted, against which brand and platform, and how it resolved — never your payload contents or secrets.

​
What each entry captures

  • Brand / platform / operation — what was targeted.
  • Outcome — ok (ran), denied (refused by scope, role, or rate limit), or error (the platform or operation failed).
  • Detail — for a denial, the reason (e.g. out-of-scope brand, read-only key attempting write, rate limit hit).
  • Timestamp and the key it came from.

​
Per-key activity

On the Settings → MCP page, open any key’s activity to see its recent calls. This is the fast way to answer “is this key actually being used, and for what?”

​
Org-wide activity

Workspace owners and admins get an org-wide view across every key in the workspace — each row annotated with the owning member’s email, the key name, and its prefix. This is your single place to answer “who is driving Mesh Pilot over MCP, and what are they touching?” Operators and viewers see only their own keys’ activity.

​
Revoke with visibility

When you revoke a key, the dialog shows its recent activity before you confirm — so you cut off access knowing exactly what that key had been doing. After revocation the key stops working immediately, but its history stays in the log.

​
What’s never logged

  • The key secret (only its prefix).
  • Payload bodies, message contents, or platform credentials.
The audit log is about accountability for capability use, not surveillance of content.

Back to Developers overview

The MCP capability model end-to-end.