Legal

Terms of Service

Effective November 10, 2026 · ProbeMyAgent — probemyagent.com

1. What this service is

ProbeMyAgent is a self-serve stress-test scanner for AI agents and chatbots. You submit an agent you own or are authorised to test, pay for a scan, and our automated system runs a fixed library of 48 test conversations ("probes") against your agent across six failure categories — prompt injection, data leakage, hallucination, tool misuse, persona drift, and refusal handling. You receive a scored report (out of 100) by email. There is no account to create and no onboarding process.

2. Authorisation — the attestation

Every intake requires you to confirm that you own the agent being tested, or that its owner has authorised you to have it stress-tested. We rely on that confirmation. You must not submit an agent, endpoint, or set of credentials that you do not own or control, or that you are not authorised to test. If we learn an intake was submitted without authorisation, we may decline to run it.

Probes are adversarial conversations of the kind a user could send your agent. Some probes ask the agent to perform actions (for example, cancelling an account). If your agent can take real actions, run scans against a staging environment.

3. Ordering, payment, and delivery

The flow is: complete the intake on this site → pay through Stripe → your scan runs automatically when payment confirms → the scored report is delivered by email (from [email protected]) to the address in your intake.
Payments are processed by Stripe. We never receive or store your card details.
Prices are in US dollars, as shown on the pricing page. The plan available for purchase at any time is the one the pricing page lists as open. Subscription plans (Watch, Fleet) are sold only when the pricing page lists them as open.
During the early launch window, report delivery can take a few hours.
Refunds. If a paid scan cannot run, or its report cannot be delivered, because of a failure on our side, you are entitled to a full refund of that scan's charge — contact us (section 7) and we will refund the original payment method. Duplicate or erroneous charges are likewise refunded in full. Once a scan's report has been delivered, the service is complete and the charge is not refundable: a low score is a test result, not a failure of the service.

4. What a score is — and is not

A score is the output of one automated probe battery at one point in time, against the agent configuration our scanner could reach. It is a testing aid, not a certification, audit opinion, or guarantee that an agent is safe, secure, or compliant. Agents change when their model, prompt, or tools change, so a past score does not describe a later version.

5. The free demo

The demo on this site is a simulation that runs in your browser. Demo results are illustrative only and are not a scan of your agent.

6. Acceptable use

Do not use the service to test systems you are not authorised to test.
Do not paste production secrets or real customer data into the intake notes. If your agent needs credentials to be reached, use read-only or staging test credentials.
Do not attempt to disrupt the service or use it for anything unlawful.

7. Contact

Questions about these terms, a charge, or a report: email [email protected]. To report abuse of this service, use the same address.

8. Changes

We may update these terms; the version published on this page at the time of your purchase applies to that purchase. Continued use of the service after an update means you accept the updated terms for later purchases.