Build a Crypto Travel Rule and AML Compliance Agent with Claude Agents SDK
Crypto-to-fiat banking access requires continuous monitoring of Travel Rule disclosures and counterparty VASP risk profiles. Build a Claude Agents SDK multi-step agent with extended thinking capabilities to audit VASP risk posture, evaluate transaction metadata, and flag suspicious mixing patterns. Achieve fully explainable AML risk scoring with structured compliance audit outputs.
What you are building
The core problem, expected build, and operating context for this challenge.
Construct an autonomous AML investigation agent using the Claude Agents SDK with extended thinking to audit Virtual Asset Service Provider (VASP) communications and compliance risk.
How work is evaluated
Evaluates the agent's ability to trace counterparty VASPs and identify missing Travel Rule originator info.
Challenge datasets
Public datasets, plus the private datasets you have access to.
How submissions are scored
These dimensions define what the evaluator checks and which criteria separate a passable run from a strong one.
verdict_enum_test
Ensures compliance_verdict matches allowable standard status strings
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
audit_accuracy
Accuracy in correctly identifying Travel Rule violations • target: 0.95 • range: 0-1
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
What you should walk away with
Build a multi-turn autonomous investigation agent using the Claude Agents SDK
Utilize extended thinking features to analyze complex multi-hop crypto transaction graphs
Parse FATF Travel Rule messaging (IVMS 101) data across counterparties
Generate detailed AML risk assessment reports for bank compliance officers
Reference links and supporting material
Dataset of 300 VASP cross-border transaction transfers containing complete and incomplete IVMS 101 Travel Rule records.
Challenge setup
Evaluates the agent's ability to trace counterparty VASPs and identify missing Travel Rule originator info.
Challenge input
JSON with transfer_id, vasp_did, ivms101_payload, and onchain_tx_hash
Claude Agents SDK
Provides extended thinking mode for complex multi-step forensic reasoning.
Evaluated output
JSON with compliance_verdict, missing_fields, risk_level, reasoning
- Ensures compliance_verdict matches allowable standard status strings
- Accuracy in correctly identifying Travel Rule violations • target: 0.95 • range: 0-1
- Benchmark: FATF Crypto Compliance Benchmark
- Audit Accuracy target: 0.95
- 1 public reference case
- Python execution harness
- Python sandbox (unavailable on Versalist)
Summary shows the first task, up to three tools, and up to three evidence items. Open the technical recipe to view all configured items.
View technical recipe
Configured tools
- Claude Code · Required
- Claude Code · Required
- Claude 4.1 Opus · Optional
- Claude 4 Sonnet · Optional
Evaluation contract
- verdict_enum_test · Weight 1
- audit_accuracy · Weight 1
Metrics
- audit_accuracy · Target 0.95
Benchmarks
- FATF Crypto Compliance Benchmark
- security
Reference cases
- 1 public reference case
Workflow tasks
- Task 1
Input: JSON with transfer_id, vasp_did, ivms101_payload, and onchain_tx_hash
Output: JSON with compliance_verdict, missing_fields, risk_level, reasoning
Execution harness: Python
Evaluation tools
- Claude Agents SDK · Agent Framework
Recipe freeze marker
No recipe freeze time is recorded.