Scan Modes
A scan mode sets how many agents Cyrion AI deploys, how deep they go and how long the scan may run. Choose it in the Scan mode row of the New Scan wizard.
A scan mode sets how many agents Cyrion AI deploys, how deep they go and how long the scan may run. Choose it in the Scan mode row of the New Scan wizard.
- Quick: 30 min
- Standard: Unlimited
- Aggressive: Unlimited
- Thorough: Unlimited
Compare modes
| Quick | Standard | Aggressive | Thorough | |
|---|---|---|---|---|
| Goal | Fast triage of the worst problems | Balanced, comprehensive testing | Deep testing | Maximum coverage |
| Time limit | 30 minutes | Unlimited | Unlimited | Unlimited |
| Agents | 2–3 | 4–6 | 6–10 | 10+ |
| Coverage | OWASP Top 10 focus, critical issues only | Broad coverage of common vulnerability classes | More specialist agents and deeper exploitation | Every vulnerability category |
| Plan | Lite and above | Lite and above | Hacker and above | Hacker and above |
Choosing a mode
- Quick — a fast sanity check after a release, or a first look at a new target. Reports critical issues only.
- Standard — the default. Right for most assessments and recurring scans.
- Aggressive — pre-release security reviews and high-value applications where you want specialists probing each area in depth.
- Thorough — the closest thing to a full manual penetration test. Use it for annual assessments, compliance evidence and critical systems.
Info: Credits are metered by actual usage, so deeper modes cost more. A Quick scan on a small site uses a fraction of a Thorough scan on a large application. See Credits & Usage.
Tune any mode
Every mode can be shaped further from the wizard.
- Max iterations caps how many steps the agents take (1 to 10,000, default 2,000).
- Max requests per second keeps load on your target predictable.
- Vulnerability focus points agents at specific categories first.
- Focus and avoid patterns steer agents toward or away from paths.
- Stealth mode routes traffic through Tor (Hacker and above).
AI Red Team attack modes
Scans of AI Application targets use attack modes designed for LLM testing. These are part of the Red Team plan.
| Mode | What it does |
|---|---|
| Single-Turn | Independent probes, each sent in isolation. |
| Multi-Turn | Conversations that escalate gradually using techniques such as Crescendo and PAIR. |
| Thorough | Runs every orchestrator for the widest coverage. |
| Adaptive | Playbook-guided follow-ups that react to what the target reveals and stop when the objective is met. |
See AI Red Team for the full picture.