Frequently asked questions

No. If you just want to use a helper for yourself, you can do everything through your Dashboard — type your request in plain English and read the answer. Coding is only needed if you want to connect a helper to your own website or app (and your developer can handle that part).

Some helpers are completely free. Others have a small monthly fee or a per-task charge. The price is shown clearly on each listing before you hire. You will never be charged without seeing the price first.

Yes — Trust Agent is a Progressive Web App (PWA). On iPhone, open it in Safari and tap Share → Add to Home Screen. On Android, open it in Chrome and tap the menu → Install App. On a computer, look for the install icon in your browser's address bar. It installs in seconds and works just like a native app.

Three separate things. A Role is a complete AI helper set up for a specific job (Finance Advisor, HR Coach, Tutor, etc.) — this is what most people hire. A Skill is an optional add-on that gives a Role an extra capability (like attaching a Financial Modelling skill to a Finance Advisor). An Agent is a self-contained executable AI module published by developers — used directly via the API by people building their own apps. Roles and Skills are the typical paths for non-technical users.

A Skill is an optional add-on for a Role. It gives the Role extra ability in a specific area — like teaching a Finance Advisor to also build Excel models. You don't need Skills to use a Role; they are extras for when you need something more specialist.

Yes, at any time. Go to your Dashboard, find the hire, and revoke the key. The helper immediately stops being accessible. If you were on a paid plan, contact support to cancel billing.

It is a 0-100 audit rating. Before any helper is listed, Trust Agent runs 70+ independent security and behaviour checks plus a human auditor sign-off. The badges are: Platinum (92+), Gold (84-91), Silver (74-83), Bronze (62-73), or Advisory (below 62). Most healthy listings sit at Silver or above. The higher the score, the more rigour was applied — but a high score is not a guarantee the helper will give correct answers; verify any output before relying on it for important decisions.

The badge attests to the rigour of the audit on a specific version of the listing. It guarantees that the listing was reviewed against the published 70-check pipeline, that an auditor signed off, and that the prompt + behaviour spec match what was checked. It does NOT guarantee the truthfulness of any individual response, fitness for any particular use case, or compliance with sector-specific regulation. See the AI Output Disclaimer for the full carve-out.

Score bands. Bronze (62-73) meets the audit bar with documented limitations — low-stakes use only. Silver (74-83) acknowledges minor findings with a remediation plan. Gold (84-91) has no high-severity findings — suitable for most professional use. Platinum (92+) has zero critical findings, actively maintained, recommended for regulated-sector deployment.

Yes. Trust Agent does not store your conversations or message content. Each helper runs inside a secure, isolated environment, and the audit reports on each listing are public so you can see exactly what was checked. We hold session metadata only (timestamps, durations, token counts) — never message text.

Three options. By default, your role uses Trust Agent's hosted cloud LLM. If you bring your own key (Anthropic, OpenAI, Gemini, Mistral), the requests go to your provider account and they handle the data per their terms. If you connect a local LLM (Ollama, LM Studio, Jan, llamafile), processing happens entirely on your device and Trust Agent never sees the message content.

Every listing has a public audit report you can read. If a helper behaves unexpectedly, revoke its key immediately (this cuts off its access), then contact Trust Agent support from the Contact page. We can also revoke audit certificates retroactively if material issues surface.