Serverless Crypto Sanctions Screening Pipeline with Claude Agents SDK and Modal
With regulatory bodies like the CFTC signaling aggressive moves toward crypto enforcement in the absence of legislation like the CLARITY Act, digital asset platforms must screen wallet addresses and smart contract interactions in near real-time against OFAC sanctions lists. In this challenge, you will build a serverless crypto compliance screening pipeline using the Claude Agents SDK hosted on Modal model runtimes. Your agent will analyze wallet transaction histories, cross-reference sanctioned wallet clusters, and issue real-time hold triggers.
What you are building
The core problem, expected build, and operating context for this challenge.
Develop a serverless sanctions screening agent using Claude Agents SDK deployed on Modal for low-latency crypto transaction analysis.
How work is evaluated
Evaluates wallet sanction identification accuracy and container execution latency.
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.
zero_false_negatives_sanctions
Ensures no known sanctioned transactions are mistakenly allowed
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
latency_p95_ms
P95 execution latency across evaluation batch • target: 450 • range: 0-2000
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
What you should walk away with
Deploy a containerized Claude Agent function onto Modal serverless infrastructure
Incorporate chain analysis tools into the agent's action loop to trace multi-hop token transfers
Achieve sub-500ms response latency for instant transaction intercept decisions
Format compliance evidence packages in accordance with CFTC and OFAC reporting formats
Reference links and supporting material
Synthetic EVM transaction graph containing simulated interactions with OFAC SDN list addresses and mixing smart contracts.
Challenge setup
Evaluates wallet sanction identification accuracy and container execution latency.
Challenge input
JSON containing transaction hash, sender, recipient, and transfer amount
Claude Agents SDK
Anthropic's official SDK supporting tool usage and reasoning.
Modal
Serverless runtime for low-latency Python execution.
Evaluated output
JSON containing action (BLOCK/ALLOW), confidence_score, and risk_factors
- Ensures no known sanctioned transactions are mistakenly allowed
- P95 execution latency across evaluation batch • target: 450 • range: 0-2000
- Benchmark: CryptoCompliance-Sanctions-Bench
- Latency P95 Ms target: 450
- 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
- zero_false_negatives_sanctions · Weight 1
- latency_p95_ms · Weight 1
Metrics
- latency_p95_ms · Target 450
Benchmarks
- CryptoCompliance-Sanctions-Bench
- security
Reference cases
- 1 public reference case
Workflow tasks
- Task 1
Input: JSON containing transaction hash, sender, recipient, and transfer amount
Output: JSON containing action (BLOCK/ALLOW), confidence_score, and risk_factors
Execution harness: Python
Evaluation tools
- Claude Agents SDK · Agent Framework
- Modal · Inference & Model Runtime
Recipe freeze marker
No recipe freeze time is recorded.