Your org's Observability dashboard
Open Dashboard in the console. It shows your plan usage (today against your daily limit, monthly API calls, total records), KPI cards for API calls, error rate, latency, Logic Blocks and agent runs, and two charts — API Calls & Errors and Metered usage — per hour for 24h and per UTC day for 7d and 30d. Pick a tenant to scope everything to it.
When you hit a limit
Once your daily limit is reached the plan strip shows 'Blocked — daily limit reached' and when it resets (00:00 UTC). Upgrading your plan, or a platform admin raising your limits, lifts the block automatically — on your next blocked request, or within about five minutes at most.
Drill-down tabs
- Tenants — calls, errors, error rate and latency per tenant.
- Resources — your busiest entities and Logic Blocks.
- Events & Logic — the event pipeline, plus a table of Logic Block and Glyph runs with errors and error rate.
- Call Log — recent calls, newest first, with filters (tenant, entity, API, status, event type), search and paging. It reads live from your logs, so a page takes a few seconds.
- Agents — your agent runs (cost, status, guardrails).
What is and isn't counted
API calls, record writes and deletes, and Logic Block runs count toward your plan limits. Glyph runs are shown but not metered, and only those run from inside a Logic Block are seen today. Published events show in the event pipeline. Traffic on public routes is not counted yet. Charts start from when counting was switched on — there is no history before that. Only average latency is tracked.
Platform admins: usage, tracing and health
Platform admins also get a cross-org Usage table (the orgs with usage in the period, against each org's real plan limits), a request & operation audit trail, and a System Health view: every scheduled platform job reports a heartbeat and records its last error, a job that goes silent for more than twice its interval shows as stale, and gateway, database and queue probes sit alongside. Request/response bodies in the audit trail are redacted in two passes before anything is stored. A separate public status page and plain /health endpoints exist for uptime monitors.