Trust & security

Verdicts, fit scores, and “not best for” notes are never influenced by vendor status, claims, or payment. The separation is structural — built into the data model, not just a policy. No one can buy a ranking.

Every table is protected by Postgres row-level security, so you can only ever read your own data. Our product analytics are first-party — captured into our own database, never sold or shared with ad networks. We don’t run third-party trackers by default.

Use your own Anthropic or OpenAI key in the Builder and it stays in your browser — it’s sent only to the model provider, never stored on our servers. You pay the provider directly with zero markup from us.

Checkout runs on Square and Stripe — we never see or store your card number. Payment webhooks are cryptographically signed and fail closed: a forged or unsigned event is rejected, so no one can grant themselves a paid plan.

Your plan and role can only be changed by a verified payment webhook or an admin — never from the browser. We hardened the database so a logged-in user cannot self-grant paid features or admin access (a common flaw we explicitly closed).

Hosted agents connect to your accounts (Gmail, Slack, Notion…) through Composio with OAuth — credentials are vaulted and scoped, and you can revoke any connection at any time. We never see your passwords.

Every hosted agent has a kill-switch, per-agent run caps, and sensitive actions (send, post, pay…) can require your explicit approval before they run. Every action is recorded in an audit trail you can review.

Agent Selector runs on US cloud infrastructure (Vercel + Supabase). We’re an independent, vendor-neutral product — your data is never sold and never used to train models, and it isn’t subject to the foreign data-handling regimes some agent vendors operate under. Export your agents and leave anytime.

We are not yet SOC 2 or HIPAA certified — and we won’t claim to be. Today we follow least-privilege access, encryption in transit and at rest, row-level security on every table, and a documented security model. SOC 2 Type II is on our roadmap as we grow; until then, this page is our commitment in writing.