Privacy Policy

Tabkeel (“we”) monitors data quality, builds a catalog and generates insights for the database you connect. This policy describes what data we process, how, and with whom.

The principles

Access to your database is read-only: Tabkeel never writes, changes or deletes anything. You connect a scoped role, choose which schemas and tables it can see, and can revoke access whenever you want.

To monitor, we run read queries and store aggregate results(counts, null rates, a metric's value), not copies of your rows. Some AI features process your schema metadata (column names and types) and, when you enable features that require it, the content you provide to that feature. This processing uses the subprocessors listed below, and you control which features you enable.

A few AI features can optionally be made more accurate by sampling your data. When — and only when — you trigger them (sample-assisted catalog descriptions and metric-history backfill), Tabkeel reads a small sample, profiles it locally, and sends the AI subprocessor a column profile only: the distinct values of low-cardinality columns and numeric or date ranges. Personally identifiable information is detected and removed locally before anything is sent, and raw rows never leave your database. The default for every feature is schema-only (names and types).

Data we process

• Account: email and name, via Clerk (authentication).
• Configuration: table and column names, checks, metrics, glossary and definitions. Metadata about your workspace.
• Metrics and scores: aggregate values computed by your checks and metrics, kept to show trends.
• Warehouse credentials: encrypted in the Supabase Vault; the app stores only a reference.
• Product usage: navigation events and, with your consent, session replay (PostHog).
• Errors: technical failure reports (Sentry).
• Payment: processed by Stripe; we do not store card data.

Security

• Traffic encrypted in transit (TLS) and data encrypted at rest.
• Warehouse credentials in an encrypted vault (Supabase Vault).
• Per-workspace isolation: a dedicated schema per organization, with RLS enabled.
• Access to your database via a least-privilege role you define.

Subprocessors

We use providers that process data on our behalf:

• Vercel — application hosting
• Supabase — database and credential vault
• Clerk — authentication
• Stripe — payments
• Anthropic — AI (Claude) for catalog, metrics and insights
• OpenAI — search embeddings (optional)
• Resend — email delivery
• Sentry — error monitoring
• PostHog — product analytics (with consent)

International transfer

Some subprocessors process data outside Brazil (for example, in the United States). When that happens, we require adequate contractual safeguards, with protection commitments equivalent to the LGPD and the GDPR.

Retention

Time series (metric and score history) follow your plan's limit, from 30 days to unlimited. When you close your account, we remove your configuration data and warehouse credentials.

Your rights (LGPD and GDPR)

You can access, correct, export or delete your data. For customers on paid plans, we provide a Data Processing Agreement (DPA) on request. Write to francisco@tabkeel.com.

Changes to this policy

We may update this policy. Material changes are announced by email or in the product, and the “last updated” date at the top always reflects the current version.

Contact

Privacy questions: francisco@tabkeel.com.