Usage and calls
Every call through every Open MCP Portal gateway in Admin › Calls, recorded contents, and the platform-wide usage dashboard.
Calls
Admin › Calls lists every call through every gateway: who called which tool, prompt or resource (a person with a key or a connected app, or a team's service account), through which gateway and server, and how it went. Platform admins and auditors read it. Filter by gateway, server, outcome, person, team, service account (any team's, searchable as "name (team)") and date.
The outcomes and the call page are the same as in a team's view; see Activity and usage. Refused calls carry the reason as their error class, for example no_access, role, gateway_disabled, key_scope, service_account, provider_disabled, team_archived or a rate-limit key.
Recorded contents
Platform admins and auditors can open the recorded contents of any recorded call from its page (View contents), after a notice that viewing is recorded. Each view writes mcp_call.payload_view to the audit log, visible to the team whose switch recorded it. Through a team's pages, platform staff never see contents. See Recording call contents.
Usage
Admin › Usage is the usage dashboard for the whole platform: calls through every gateway, how they went and how long they took.
- Range: last 24 hours, 7, 30 or 90 days, or custom dates in UTC (at most 90 days; hourly grouping at most 7).
- Group by: gateway, server, capability, team (platform gateways in one row), kind of caller, day or hour. Team narrows to one team: its gateways, its servers and its service accounts. There is no grouping by person.
- Totals, calls per day by outcome, p50 and p95 latency, and bytes sent and received.
Usage comes from hourly rollups, recomputed every 5 minutes on one replica, so the dashboards lag by a few minutes. Usage rollups are kept for 400 days by default, longer than call records.