Skip to content
Telemetry

Observe the data

This guide explains how to observe a Supabase project.

Use this page to:

What data you can observe #

The sources are logs, metrics, live Postgres diagnostics, and advisor findings. Open Logs and Reports in Studio when you want a UI on those sources. See where you can observe it.

Logs#

Request, database, Auth, Storage, Realtime, and function events in ClickHouse.

Metrics #

Prometheus-compatible CPU, IO, WAL, connections, and query stats. Scrape the Metrics API for custom dashboards, alerting, or retention beyond Studio.

Database#

Live Postgres statistics such as bloat, cache hit rate, blocking sessions, index usage, and slow queries. See Inspect the database (supabase inspect db) for the command and SQL catalog.

MCP execute_sql can run the same read-only queries.

Advisors#

Deterministic security and performance findings. Pull them from Advisors.

Where you can observe it #

Use the interface that matches where you are working.

MCP #

Configure the Supabase MCP server for one project with read-only mode. The debugging tools are:

  • query_logs for bounded log queries. Use the same ClickHouse SQL as Query and filter logs.
  • execute_sql for read-only database inspection
  • get_advisors for security and performance findings

Ask the agent to inspect a time window, error code, or advisor finding. Keep production connections project-scoped and read-only.

API #

Use the Management API when you want the same data from a script:

Scrape the Metrics API for Prometheus-compatible database series. That endpoint is a project URL, not a Management API route.

CLI #

Use the Supabase CLI against a linked project:

Run supabase inspect db help on the installed CLI to see the reports available in your version.

Studio #

Use Studio when you want to inspect the project in the browser:

  • Logs for the unified Logs view
  • Reports for API, Auth, Storage, Realtime, and database dashboards
  • Advisors for deterministic security and performance findings

To send logs or traces to your own stack, see Log drains, Client-side tracing, and Sentry integration.