Before your AI agent acts, test what its approval actually proves.
A concrete, one-time assessment of one high-impact agent workflow — payments, refunds, customer records, API writes or deployments. Three controlled failure scenarios and a written evidence report. €299. No subscription.
What you receive
- 01 — Action-to-authority mapDocument the operator's intended action, the delegated agent, arguments, limits and exact approval boundary.
- 02 — Three synthetic negative testsCheck (1) mismatched recipient, amount or destination; (2) changed action arguments after approval; (3) a lost response or retry whose downstream effect is uncertain. Exact fixtures are scoped to your workflow.
- 03 — Reproducible evidence reportExpected vs. observed behavior, PASS / FAIL / UNKNOWN per test, public or redacted reproduction steps, residual risks and next actions. If independent outcome evidence is unavailable, the result is UNKNOWN — not a claimed success.
The audit uses synthetic or explicitly authorized test data. We do not ask for wallet private keys, payment authorizations, real client records or production credentials. The assessment is AI-assisted and is independent of the target's implementation; it is not a human-led penetration test, insurance opinion, security certification or proof of regulatory compliance.
Technical evidence you can check before paying
These are public product artifacts — not customer endorsements or claims that a particular client's workflow has been audited.
- Example: executable checks for evidence fidelity, unknown outcomes and unsupported revenue claims
- Machine-readable API contract for advisory Safe Preflight and execution evidence
- x402 endpoint — payment challenge can be inspected without paying
Frequently asked
Does this prove my agent is secure?
No. It evaluates a tightly bounded scenario. The report separates verified behavior from unknown or out-of-scope risks.
Do you deploy code or change my permissions?
Not in this €299 audit. The separate €990 Production Pilot can scope an integration, subject to technical feasibility and agreed access.
Will I need to provide real wallets or confidential customer data?
No. We work from synthetic fixtures, public documentation, or specific artifacts you are authorized to share.
What if my workflow cannot be evaluated from synthetic fixtures?
We clarify feasibility and an acceptable alternative scope before substantive work. Do not upload secrets to the checkout or email.