Put boundaries around production work

Captor runs inside your Node.js application. Declare a resource ceiling, reserve capacity before side effects, record progress after successful work, and verify the outcome. Keep your existing scheduler and database client.

Start with the quickstart, then run the recovery demo. No Captor account, database, provider key, or hosted service is required for either.

• Execution contracts: resources, reservations, metrics and deadlines.
• Backfills and resume: batches, durable progress and replay boundaries.
• Stores and CLI: inspect local receipts and failures.
• Adapters: bounded fetch calls and Prisma writes.
• SDK API: public execution API.
• Platform receipts: optional manual import and inspection.

The product is Captor; the npm package is captar. The existing OpenAI-compatible runtime remains supported through the AI compatibility quickstart.