Trust & security
What we do today, stated plainly. Where a certification is not yet held, we say so.
Data isolation
- Every company runs in its own workspace (
houses/<id>) with its own agents, knowledge and usage ledger.
- Tenant isolation is tested automatically; a user scoped to one company cannot list or reach another company's agents.
- Each agent is restricted to its own filesystem scope; command execution is disabled fleet-wide.
Transport & secrets
- TLS 1.2+ on all public endpoints (TLS 1.3 negotiated where supported), HSTS enabled.
- Credentials are stored masked; the model never receives secret values, and secrets are never printed to logs.
- Provider keys are scoped per provider and can be rotated without downtime.
Subprocessors
- Z.ai — primary model inference (GLM 5.3 Flash)
- MiniMax — secondary model inference (fallback only)
- PayTabs — payment processing (Saudi Arabia)
- Cloudflare — DNS, TLS termination and edge protection
Data retention
- Conversation and audit records: retained for the life of the workspace, deletable on request.
- Operational logs: 90 days.
- Deletion requests are executed within 30 days, with a written confirmation.
Compliance status — honest
SOC 2: readiness programme in progress — not certified.
GDPR / PDPL: alignment in progress — data export and deletion are implemented; a formal DPA is available on request.
Penetration testing: internal black-box review completed; an external engagement is planned before enterprise launch.
Reporting a security issue
Contact the platform owner directly. We acknowledge reports within 2 business days and will not pursue action against good-faith researchers.
This page describes controls that are implemented and verifiable today. It is updated when a control changes.