Ownership, with clear boundaries

Understand where your data goes

Your gateway and PostgreSQL database run on infrastructure you choose. Cloud models, voice APIs, connected apps, MCP servers and web tools may receive data when used. Local inference is available, but enabled tools can still make network requests.

This website is a static product site. It does not connect to your OwnPilot instance or collect provider API keys. The theme preference is stored in your browser; no analytics or advertising scripts are included.

Authentication is a deployment decision

OwnPilot supports none, API-key and JWT authentication modes. The security policy describes localhost as the default trust boundary. Set API authentication and a UI password before exposing the gateway; use HTTPS at your reverse proxy.

A configured API key without a UI password is not the same as securing every browser surface. Follow the policy’s web UI and WebSocket guidance, and review allowed origins.

Sensitive configuration and encryption

The product documents AES-256-GCM cryptography and encryption of stored service configuration. OWNPILOT_ENCRYPTION_KEY can supply key material; otherwise a data-encryption key is generated in the data directory. Preserve that key with your backup plan.

ENCRYPTION_KEY is separately documented for OAuth-token encryption in production. These controls do not establish that the entire database is encrypted, and they are not end-to-end encryption for requests to external providers.

Tools, approvals and isolation

Documented code-execution controls include critical-pattern checks, permissions, an approval callback and sandbox isolation. Docker and local execution have different trust boundaries. CLI tools also have allowed, prompt and blocked policies.

External skills and extensions execute code or instruct an agent. Review capabilities and sources, use narrow permissions and keep host-access opt-ins disabled unless needed. No sandbox or automated audit is a guarantee that arbitrary code is safe.

Audit and network controls

The project documents hash-chain audit verification, rate limiting, PII detection/redaction and SSRF protections. Configure and review these controls for your deployment; their presence is not a claim of a third-party security certification.

Report a vulnerability

Follow SECURITY.md in the product repository for the reporting procedure and scope. Do not include credentials, personal data or an exploit against someone else’s deployment in a public issue.