Intelligence Reference
V3 Intelligence Sources & Risk Categories
Radar V3 aggregates intelligence from multiple proprietary and external sources, evaluating addresses across 8 primary risk categories with 52+ signal types.
Intelligence Sources
Proprietary Intelligence
Real-time threat actor monitoring
Historical transaction pattern analysis
Cross-chain entity resolution
Machine learning risk models
External Data Sources
Regulatory sanctions lists (OFAC, EU, UN)
Stablecoin issuer blacklists (USDT, USDC)
Security threat feeds
Exchange compliance data
On-Chain Analysis
Transaction pattern detection
Address poisoning alerts
Sybil attack detection
Suspicious funding patterns
Source Anonymization
All API responses use codenames instead of actual data source identifiers to protect our intelligence methodology.
| Codename | Description |
|---|---|
alpha | Primary blockchain intelligence |
bravo | Security threat monitoring |
charlie | Smart contract analysis |
delta | On-chain analytics |
echo | Market data aggregation |
foxtrot | Market intelligence |
golf | Asset analytics |
hotel | Stablecoin blacklist detection |
india | Auto-whitelist system |
juliet | AI calibration engine |
Note: Codenames may change without notice. Do not hard-code logic based on specific codename values.
Risk Categories
V3 scoring model evaluates 8 primary risk categories with 52+ signal types, organized into tiers by severity.
Tier 1: Critical
Sanctions matches (OFAC, EU, UN)
Stablecoin blacklist status
Known threat actors (hackers, scammers)
Attack detector alerts
Tier 2: High Risk
Mixer/tumbler interactions
Suspicious funding sources
High-risk counterparties
Ransomware associations
Tier 3: Medium Risk
Unknown exchange interactions
Privacy service usage
Large cash-out patterns
Bridge protocol activity
Tier 4: Low Risk
Minor risk indicators
Low transaction volume
New address with limited history
Unverified but non-suspicious activity
Error Handling
All errors follow a consistent JSON format:
{
"error": "Validation Error",
"message": "Invalid address format",
"details": {
"field": "address",
"received": "invalid123"
},
"timestamp": "2025-01-06T10:30:00.000Z"
}HTTP Status Codes
| Code | Description |
|---|---|
| 200 | Success |
| 400 | Bad Request - Invalid input |
| 401 | Unauthorized - Invalid/missing API key |
| 429 | Rate Limited - Too many requests |
| 500 | Internal Error - Service issue |
| 504 | Timeout - Processing exceeded limit |
MCP Server Integration
Radar V3 is designed for seamless integration with Model Context Protocol (MCP) servers and AI agent frameworks.
Response Design for MCP
Flat Structure: Top-level fields for easy parsing
Consistent Types: Predictable field types across responses
Action-Oriented: Clear recommendation.action for decisions
Contextual Summary: AI-generated summaries for agents
Example MCP Tool Definition
{
"name": "check_address_risk",
"description": "Check blockchain address risk score and compliance status",
"inputSchema": {
"type": "object",
"properties": {
"address": {
"type": "string",
"description": "Blockchain wallet address to check"
}
},
"required": ["address"]
}
}Recommended Integration Pattern
// MCP Tool Handler
async function handleAddressCheck(params) {
const response = await fetch('https://api.radar.getfailsafe.com/api/v3/intel', {
method: 'POST',
headers: {
'Authorization': `Bearer ${API_KEY}`,
'Content-Type': 'application/json'
},
body: JSON.stringify({ address: params.address })
});
const data = await response.json();
// Return structured response for LLM
return {
risk_score: data.score,
risk_level: data.risk_level,
action: data.recommendation.action,
summary: data.recommendation.summary,
is_sanctioned: data.flags.sanctioned,
is_blacklisted: data.flags.blacklisted
};
}Best Practices
Performance Optimization
- • Use batch endpoint for multiple addresses
- • Cache high-risk scores (60+) for 10 minutes
- • Only request include_details when debugging
- • Use batch_size: 3 for optimal throughput
Compliance Integration
- • Log correlation_id for audit trails
- • Default to REVIEW action on timeouts
- • Implement exponential backoff on 429
- • Use /feedback to improve model accuracy