🔬

Agent Scanner

Antivirus for AI agent responses. Real-time scanning for prompt injection, hallucination, data exfiltration, jailbreaks, and PII leakage.

48.2K

Scanned / 24h

Detection Types

<50ms

Scan Latency

99.2%

Detection Rate

Get Free API Key 🔑

All data shown is simulated for demonstration purposes. Live data requires an API key.

48,219

Responses Scanned

Injections Caught

156

Hallucinations

47ms

Avg Latency

Recent Scan Results

🚨 Prompt Injection Detected

Agent output contained embedded system override instructions targeting downstream consumers.

Critical

2 min ago

💡 Hallucination Flagged

Financial data claim doesn't match any known source. Confidence: 12%.

High

8 min ago

📤 Data Exfiltration Attempt

Agent response included encoded PII in base64 payload. Auto-blocked.

Critical

15 min ago

✅ Clean — No Threats

Response passed all 6 detection categories. Safe to deliver.

Safe

18 min ago

Detection Categories

Category	Method	Severity	24h Count
Prompt Injection	NLP classifier + pattern matching	Critical	23
Hallucination	Fact-checking + confidence scoring	High	156
Data Exfiltration	Regex + semantic analysis	Critical	7
Jailbreak Patterns	Adversarial prompt detection	Critical	4
Toxic Content	Content moderation classifier	High	31
PII Leakage	Named entity recognition	High	12

Combine Agent Scanner with Agent Firewall for a complete defense-in-depth strategy.

API Example

            
            const result = await ac.scanner.scan({
  agentDid: 'did:agent:abc123',
  output: 'Here is the financial report...',
  context: 'customer_support'
});

console.log(result.safe);       // true or false
console.log(result.threats);    // [{ type, confidence, segment }]
console.log(result.action);     // 'allow' | 'flag' | 'block'
        