AgentPay Sentinel MCP 🔐

Autonomous security guardrails for AI agent payments. Red team attack simulation + blue team defense = adversarial co-evolution.

Part of the AgentPay implementation layer — the governance fabric that makes agent payments enterprise-grade.

What It Does

AgentPay Sentinel runs 9 security checks on every agent transaction:

| # | Check | What It Blocks | |---|-------|---------------| | 1 | Token Integrity | Forged/replayed scoped tokens | | 2 | Budget Enforcement | Spend exceeding budget cap | | 3 | Merchant Allowlist | Payments to unauthorized merchants | | 4 | Category Restriction | Purchases in blocked categories | | 5 | Expiry Check | Expired tokens | | 6 | Rate Limiting | Transaction velocity abuse | | 7 | Amount Verification | Amount changed after approval | | 8 | Replay Detection | Duplicate transactions | | 9 | Chain Integrity | Tampered audit logs |

Tools

• sentinel_audit_transaction — Full 9-check audit with SHA-256 hash
• sentinel_verify_chain — Verify audit chain integrity
• sentinel_threat_model — Red-team simulation against guardrails

Inspired By

Project ARES — autonomous adversarial security operations. AgentPay Sentinel applies the same red-team/blue-team co-evolution pattern to payment governance.

Pricing

• Free: 50 audits/month
• Pro: $19/mo unlimited audits + threat modeling