Security
Security built into every action
Agents that can act need stronger controls than chatbots. Here is exactly how Papple AI protects your workspace.
Controls
Tenant isolation
Every record carries a workspace ID and every query is constrained by it at the data-access layer. There are no ID-only lookups.
Encrypted secrets
Integration credentials are encrypted at rest with authenticated encryption (libsodium) and are write-only: they can be used by tools but never read back through the app or API.
Authentication
Strong password policy, optional TOTP two-factor authentication with replay protection, email verification and brute-force rate limiting.
Least-privilege roles
Owner, admin, builder, approver and viewer roles. API keys are additionally limited by scopes and cannot approve actions.
Safe tool execution
Consequential actions require human approval by server-side policy. Only the exact approved payload executes. Model output can never bypass it.
Outbound request protection
The HTTP tool allows https only, blocks private and reserved IP ranges, disables redirects and caps response size to prevent SSRF.
Rate limits
Per-user, per-key, per-workspace and per-IP limits on sign-in, API calls, runs, webhooks, forms and outbound email.
Audit log
Hash-chained, append-only audit entries for sign-ins, configuration changes, runs, approvals and tool executions, with one-click integrity verification.
Prompt-injection hygiene
User-supplied content is treated as untrusted data in prompts; tools are chosen by the workflow definition, not by model output.
Questions from your security team?
Enterprise plans include a security review, custom data retention and SSO.