Before an Agent Touches Money,
It Calls Us.
8-step security pipeline for AI agents. Trust attestation, behavioral fingerprinting, capability escalation, adversarial honeypots. One API call. Under 10ms.
89 endpoints. Pay per call with USDC on Base via x402. No subscription required. Your agents authenticate autonomously.
2,000 Agents. 100K Screenings/Day.
Picture a quant fund running 2,000 AI agents, each executing 50 trades per day. Every trade screened through our 8-step pipeline before execution. 100,000 screenings per day. Each agent builds a verifiable trust chain that proves which agent authorized which trade, with a signed head for the entire chain. This is what MerchantGuard was built for.
8-Step Security Pipeline
Every agent call passes through all 8 steps. Total latency under 10ms.
Sliding window rate limiting per agent ID. Prevents brute-force trust escalation.
Behavioral fingerprinting for cross-platform correlation, no plaintext identifier exchanged.
L1 Redis cache with L2 Supabase fallback. Block keys for agents with trust below 50.
Maps requested action to 4-tier capability levels. Trust score must meet threshold.
Suspicious agents routed to decoy endpoints. Adaptive response delays neutralize probe-and-pivot attacks.
Transcript reconciliation separates dialect variation from adversarial manipulation.
Perceptual binding with cryptographic signature. Replay protection on 5-minute window.
Trust attestation issued. Audit chain head signed. Capability escalation unlocked.
Capability Escalation
Agents earn higher capabilities as trust grows. New agents start at READ_ONLY.
READ_ONLYQuery data, read public endpoints
DATA_WRITESubmit data, update profiles
PAYMENT_INITIATECreate payment intents, initiate transfers
PAYMENT_EXECUTEExecute transactions, manage funds autonomously
Quick Start
Endpoint Reference
89 v2 endpoints across 7 categories. All authenticated via sk_live_ API key or x402 USDC.
POST/api/v2/agent/screen8-step screening pipeline. The core endpoint.
POST/api/v2/agent/registerRegister an agent identity. Returns agent ID + webhook URL.
POST/api/v2/agent/guardgateVerify agent trust before granting capabilities.
GET/api/v2/agent/verify-dispatchVerify an agent dispatch token.
x402 USDC Pricing
AI agents pay per call with USDC on Base. No API key. No subscription. No human in the loop.
Also available via API key: Free (1,000 calls/mo) · Solo $19/mo · Enterprise $299/mo
The Math: Build vs. Buy
Trust Attestation
When agents work together, trust must flow through the chain. Our directed acyclic graph tracks fan-in and fan-out trust relationships with cryptographic verification at every node.
Topological Sort
Kahn's algorithm validates the trust chain has no cycles. Circular trust dependencies are rejected.
Merkle Root
SHA-256 Merkle tree over the entire trust chain. One hash proves the integrity of every agent in the graph.
Fan-In / Fan-Out
Multiple agents can contribute trust to a single node (fan-in). A trusted agent can delegate to multiple downstream agents (fan-out).
Capability Gates
Each node in the DAG has a minimum trust threshold for its capability level. The chain is only as strong as its weakest link.
Authentication
API Key
Get an sk_live_ key from the developer portal. Include as Authorization: Bearer header or x-api-key header.
Free: 1,000 calls/mo. No credit card.
x402 USDC
No key needed. Include X-PAYMENT header with amount. USDC on Base L2. Sub-cent gas. Instant settlement.
For autonomous agents that can't sign up.
The Compliance Layer for the Agent Economy.
One API call. Every agent verified. Every trust chain auditable.
Protected by U.S. patent-pending technology (App. Nos. 63/983,615; 63/983,621; 63/983,843; 63/984,626). Additional patents pending.

