Clear answers about the meter.

No. DatabaseMeter accounts and dashboard access are free to create and use, and no credit card or payment information is required. Neon, Supabase, or another PostgreSQL provider may still charge for its own service and usage; scheduled telemetry can create ordinary connection or compute activity at that provider.

DatabaseMeter supports Neon and Supabase through their Management APIs and can monitor compatible PostgreSQL 14–18 databases through a dedicated read-only connection. Provider availability still depends on the credential scope, plan, enabled APIs, and values each upstream service returns.

Yes. A user-owned portfolio can contain multiple Neon and Supabase connections plus direct PostgreSQL projects. Discovered projects can be monitored or paused independently, so an accessible provider project does not have to appear in the active portfolio.

The database must be publicly reachable on a PostgreSQL TLS port and use a dedicated least-privilege monitoring role. DatabaseMeter requires certificate-verified TLS, validates the destination server-side, and does not accept private or mixed DNS targets, Unix sockets, multiple hosts, routing overrides, or client certificates.

The guided setup creates a dedicated login with CONNECT, a read-only default, a small connection limit, and statistics visibility supported by the provider. pg_stat_statements contributors also require the extension and sufficient cross-role statistics visibility; managed-provider permission rules can limit what is available.

Provider and PostgreSQL telemetry collection use separate protected schedulers. PostgreSQL telemetry can be collected manually or scheduled per database for hourly, every six hours, or daily collection. A connection can be paused without deleting its stored history.

The dashboard combines a portfolio overview with database detail, storage, compute, cost-planning, alert, and synchronization views. Available data can include size and growth, provider usage, endpoint or resource evidence, relation statistics, connections, and aggregate workload counters.

No. Provider usage describes metered consumption, provider resources describe infrastructure configuration or gauges, and PostgreSQL telemetry describes database-visible activity. DatabaseMeter keeps these sources separate and does not turn them into a fabricated common score.

No. Observed SQL execution, reads, writes, sessions, and wait evidence can help investigate workload, but they do not assign an exact share of CU-hours, compute-hours, or a provider bill to a query unless the provider supplies an authoritative mapping.

DatabaseMeter displays it as unavailable and explains provider or capability limits where possible. A missing value is never silently replaced with zero, copied from an unlike metric, or estimated without a stated model.

Qualified Neon planning ranges can use compatible stored usage and versioned rules. DatabaseMeter does not present those ranges as an invoice and does not fabricate Supabase or direct PostgreSQL costs when compatible usage and allocation evidence are absent. Public calculators remain scenario tools based on values you enter.

In-app rules evaluate only when their required stored metric exists. Weekly reports are an explicit opt-in and summarize noteworthy changes, unresolved findings, and quiet projects from data already stored by DatabaseMeter; generating a report does not make new provider calls.

A signed-in user chooses a predefined operational question and one database, all monitored databases, all Neon, all Supabase, or all other monitored sources. DatabaseMeter then assembles only the bounded monitoring evidence relevant to that question and requests a structured, evidence-linked interpretation from OpenAI. The result is advisory and cannot execute database or provider changes.

The pre-run disclosure shows the selected target count and monitoring categories. DatabaseMeter excludes credentials, connection strings, hostnames, role names, authentication identifiers, SQL text, application rows, and raw provider responses. OpenAI response storage is disabled for the request. DatabaseMeter keeps the validated result and safe run metadata for 30 days unless the user deletes it sooner; the full input package is not retained. Requests are rate limited.

Yes. The public tools include a PostgreSQL health check, a pg_stat_statements analyzer, database-growth and provider-usage calculators, PostgreSQL size references, and a Neon-versus-Supabase comparison. The PostgreSQL Learn Center is also public.

No. Calculator inputs and results stay in your browser. They do not save values, connect to a provider account, or turn a scenario into a billing promise.

Visitors who are not signed in can paste bounded exports for local analysis. A signed-in user can instead choose an eligible owned telemetry connection and run the same fixed analysis without copying data. One-time reports are not persisted, and credentials never return to the browser.

No. The analyzer rejects or discards SQL-text columns and works with bounded aggregate counters and query fingerprints. Connected analysis uses fixed server-side queries that exclude SQL text.

The Learn Center contains source-backed guides about database and index size, query and session statistics, sequential scans, growth, dead tuples, autovacuum, WAL, connection pooling, idle transactions, cache and I/O, and managed PostgreSQL monitoring. Each guide distinguishes what a signal shows from what it cannot prove.

No. Normal PostgreSQL monitoring uses fixed bounded queries against system catalogs and aggregate statistics. It does not select application tables, store customer row contents, run mutations, request EXPLAIN ANALYZE, or collect SQL text.

Auth0 handles authentication. Provider and PostgreSQL credentials are validated server-side, encrypted with AES-256-GCM under a separately versioned environment key, decrypted only for the required server operation, and never displayed after submission. Ownership is checked server-side for user-specific resources.

DatabaseMeter retains connected-project metadata, supported monitoring snapshots, synchronization history, alert state, settings, and opted-in email delivery history. Browser-local calculator inputs and one-time health-check or analyzer reports are not persisted.

Disconnect controls remove individual DatabaseMeter connections. Account deletion permanently removes DatabaseMeter-owned credentials, projects, snapshots, alerts, settings, email delivery history, and sync history. It does not revoke keys at Neon or Supabase and does not delete the upstream Auth0 identity, so external credentials should be revoked separately.

No. Demo mode is an explicit preview built from deterministic synthetic projects and is labeled separately from live connected data.

No. Portfolio views, metrics, trends, deterministic cost-planning rules, alerts, tools, and reports continue to work without AI analysis. AI analysis is an optional on-demand interpretation layer and may be temporarily unavailable during service maintenance.