Where Does Your Schema Go?

Short Answer: Nowhere. Schemap runs entirely on your local workstation or CI/CD runner. It only reads database structural catalogs (tables, columns, foreign keys) and writes compiled Markdown directly to your project repository. Your database row data, customer PII, passwords, and records are never read, dumped, or transmitted.

Schemap connects to your database using standard SQL driver connections (PostgreSQL, SQLite, MySQL, Turso, Oracle), but it never issues SELECT * or queries customer data tables.

It exclusively inspects read-only system catalog metadata:

-- What Schemap queries (Metadata only):
SELECT table_name, column_name, data_type, is_nullable
FROM information_schema.columns
WHERE table_schema = 'public';

-- What Schemap NEVER queries:
SELECT * FROM users;          -- ❌ NEVER EXECUTED
SELECT email, ssn FROM customers; -- ❌ NEVER EXECUTED

All core CLI commands (schemap context, schemap doctor, schemap join, schemap agents, schemap diff) execute using local Python algorithms—including foreign key graph traversal, centrality ranking, and safety rule generation. No generative AI or cloud endpoints are queried during normal compilation.

Schemap contains no usage tracking, no Google Analytics, no PostHog telemetry, and no behavioral logging. Your schema names, column structures, and query patterns remain strictly confidential to your git repository.

When compiling agent rules (CLAUDE.md, AGENTS.md, Cursor rules), Schemap actively searches for sensitive columns (e.g. password_hash, api_secret, stripe_token, ssn) and injects explicit negative guardrails instructing coding agents never to query them in feature SQL.

How Schemap compares against other database tools in the AI ecosystem:

To ensure complete transparency, there are only two circumstances where network communication can ever occur:

• Use Read-Only Credentials: When configuring DATABASE_URL, provide a database user with read-only access to system catalogs (e.g., GRANT USAGE ON SCHEMA public).
• Commit Context Maps to Git: Commit schemap_database_context.md, CLAUDE.md, and AGENTS.md directly into your repository. Your team members can consume full database intelligence without needing direct database access.
• Run in CI/CD Pre-Commit Gates: Use schemap diff --risk --fail-on-breaking in your automated deployment pipeline to catch destructive schema changes before they hit production.