PlanGate
Type to search every docs page
↑ ↓ to move↵ to openesc to close
DocsSign in

Security

Last updated: November 9, 2026

Your AI requests

Prompts and responses pass through PlanGate to the app's AI provider and back, and are not stored. For each request we keep only the model, the token counts and the cost, never the content.

PlanGate trains no models and gives request contents to nobody but the app's provider. Server logs and error reports never contain request or response bodies.

Keys and credentials

  • Provider keys, payment tokens and store credentials are encrypted at rest with AES-256-GCM, with a key kept outside the database. They are never sent back, not even to their owner.
  • A provider key is decrypted only in memory, to call that provider for that app's users.
  • We recommend giving PlanGate a key of its own, with a spending limit set at your provider (an OpenAI project budget, an Anthropic workspace limit…), so that a leaked key could never cost more than that limit.
  • Sign-in tokens are stored only as hashes, and expire.
  • Staging runs on its own database with its own encryption key and never holds production data.

Access and isolation

  • Every request is checked against its owner: a developer sees only their own apps and customers, an end user only their own usage.
  • Each end user's spending is reserved before a request is sent, so parallel requests can't exceed a budget, and each caller is rate-limited.
  • All traffic uses HTTPS.

Availability and backups

  • The database is backed up continuously by Neon, with point-in-time restore.

Reporting a vulnerability

Write to [email protected] with "Security" in the subject. We answer within 3 business days and keep you informed until it is fixed.

We will not take legal action against research done in good faith that:

  • respects these rules,
  • avoids harming users, their data and the service (no denial of service, no spam),
  • uses only accounts you own,
  • gives us reasonable time to fix the issue before you disclose it.

Our security.txt follows RFC 9116.

❌🕸️