LintBase is positioned as the "ground truth for AI coding agents," aiming to solve the problem of AI tools (such as Cursor, Windsurf, Copilot Workspace, and Claude Code) generating incorrect code because they do not understand the real database schema. It connects directly to the database, reads the real structure of live documents, and generates AI-optimized context.
Main capabilities include:
- Preventing AI hallucinations: generates precise schemas, field occurrence rates, and types.
- Catching schema drift: compares against schema snapshots via the lintbase check command, providing protection in CI.
- Safe context: highlights missing rules or exposed PII before AI writes queries.
- Cost awareness: prevents AI from writing unbounded queries on collections with more than 2 million documents.
- General NoSQL: supports Firestore and MongoDB.
How to use:
- Get a service account key: Firebase Console → Project Settings → Service Accounts → Generate New Private Key, save the JSON file, and never commit it to git.
- CI pipeline protection (schema drift): run the snapshot command to create a baseline, commit .lintbase/schema.json to the repository, and then add the check command to the CI/CD pipeline. If a query or deployment accidentally deletes a critical field or changes a type, the CI build fails immediately.
- Run a general scan: view the full report in the terminal.
- Save to the dashboard (optional): track database health at lintbase.com; you need to obtain a token at lintbase.com/dashboard/settings.
Supported database commands:
- Firestore: npx lintbase scan firestore --key ./sa.json
- MongoDB: npx lintbase scan mongodb --uri mongodb+srv://user:pass@cluster.mongodb.net/test
AI agent integration (MCP): When using Cursor, Claude Desktop, or Windsurf, installing lintbase-mcp provides AI agents with real-time Firestore schema context. AddAfter adding it to .cursor/mcp.json, when you ask AI to "add a field to users," it will check the real schema before writing code.
Applicable scenarios:
- Developers using AI coding tools to build Firestore or MongoDB applications.
- Teams that need to prevent AI from generating incorrect database queries.
- Projects that want to detect schema drift in CI.
- Scenarios that need to understand database security rules and cost risks.
Limitations:
- Only Firestore and MongoDB are supported; Postgres is an obvious gap and will be supported soon.
- Large collections use sampling rather than full scans, and occurrence rates are estimates on collections with millions of documents.
- The MCP server is newer than the CLI and may encounter rough edges first.
Pricing: The CLI local tool is 100% free; the Pro version is $39 per month and provides unlimited history, a dashboard, and shared team workflows. Service account keys never leave the local machine and are only read locally; document sampling has a hard limit of --limit (default 100) to prevent accidental read costs; the --save flag only sends scan summaries and issue lists, never raw document data.
