FILE № AUDIT · FREE SECURITY SCAN

Find out what your agent does wrong.

RAE runs 50–200 adversarial prompts across 10 attack categories against your live endpoint. Severity-ranked findings in five minutes. No account. No credit card.

50–200

Attacks per audit

Attack categories

5–10 min

First findings

On-prem

Your prompts stay yours

HOW IT WORKS

Submit your work email and HTTPS endpoint

Paste your agent's live endpoint URL and your work email address. Personal emails are not accepted.

RAE attacks it — async

50–200 adversarial prompts across 10 categories run against your endpoint. No human in the loop. Results arrive in 5–10 minutes.

Get your severity-ranked findings report

Critical to low — every finding ranked and actionable. Full debrief on a 30-min call.

Personal email addresses (Gmail, Yahoo, Outlook) are not accepted.

START YOUR AUDIT

Free. No account. No credit card. Results in 5–10 minutes.

10 ATTACK CATEGORIES

PROMPT INJECTION

JAILBREAK

DATA EXFILTRATION

ROLE CONFUSION

TOOL ABUSE

SCOPE DRIFT

MULTILINGUAL SWITCHING

IDENTITY SPOOFING

SYSTEM PROMPT LEAK

CONTEXT MANIPULATION

FIG. 02 — Attack surface map · 10 categories · 50–200 prompts total

DELIVERABLES

A dossier on your
agent's vulnerabilities.

Every audit produces four structured outputs. No dashboards. No portals. A clean report delivered to your inbox, then walked through live.

SEVERITY RANKING

Every finding ranked

Critical, high, medium, and low — prioritised so you know what to fix first.

ATTACK PROMPTS

The exact prompts that succeeded

The full set of adversarial inputs your agent failed to handle correctly.

RESPONSE LOG

What your agent said back

Verbatim responses captured for every finding — no sanitisation, no redaction.

HARDENING BRIEF

What to fix first

Actionable remediation steps, delivered and walked through on a 30-min call.

Dossier № Audit · No account · No credit card · 5–10 min