VAMP Enforcing · 0.9% threshold live
Free VAMP Calculator
Breaking
Faceless YouTube "Automation" courses · Stripe reviewing accounts <3mo history(2h ago)·CRITICAL: Vulnerability in popular AutoGPT fork allows RCE via prompt injection. Patch immediately.·Visa VAMP enforcement ramps to 0.9%. issuer pre-flagging hits 4× normal rate(4h ago)·Stripe Sessions Apr 29. agentic checkout panel confirms compliance "is the bottleneck"·Coinbase x402 ecosystem listing live. agent-native payments shipping at scale·Faceless YouTube "Automation" courses · Stripe reviewing accounts <3mo history(2h ago)·CRITICAL: Vulnerability in popular AutoGPT fork allows RCE via prompt injection. Patch immediately.·Visa VAMP enforcement ramps to 0.9%. issuer pre-flagging hits 4× normal rate(4h ago)·Stripe Sessions Apr 29. agentic checkout panel confirms compliance "is the bottleneck"·Coinbase x402 ecosystem listing live. agent-native payments shipping at scale
×
The Compliance Layer for the AI Agent Economy

AGENTGUARD

The compliance layer for the AI agent economy.OFAC sanctions · LLM firewall · Kill Switch · sub-50ms verdict per call.

Quick start · Claude Code
$
claude mcp add guard --transport sse https://merchantguard-mcp-810654658669.us-central1.run.app/mcp/sse
51tools instantly available · sub-50ms verdict per call
See architecturePay-per-call · USDC
AgentGuard™ USPTO 99462472OFAC SDN screening · liveSub-50ms · USDC on Base
Are you a merchant?See the merchant funnel →
TAGLINE
BUILT ON
Stripe / Visa / Mastercard / OFAC / FATF / EMVCo / x402 / ERC-8004

Today, AI agents are starting to move money. They book trips, pay invoices, subscribe to services, dispute charges. But there's no portable way to verify whether a specific agent action was authorized, signed by a real human, or compliant with the rails it's touching.

Credentials leak. Sessions get shared. Chargebacks come back. Downstream systems can't tell trusted agents from malicious ones, so they either block both or trust both.

Until now.

Verify any agent action in 4 lines.

A signed compliance verdict on every agent transaction. OFAC, LLM firewall, and Kill Switch in one API call. Sub-50ms.

// Pull the agent action context
const action = await agent.getPendingAction()

// Compile a signed compliance verdict
const verdict = await agentguard.verify(action)

// Branch on the cryptographically signed result
if (verdict.allowed && verdict.signature) {
  agent.execute(action, { audit: verdict.attestation })
}
▸ verdict signed in 47ms
One API call · Sub-50ms

Six layers. One signed verdict.

Agent inputs funnel into the MerchantGuard® compliance layer (KYA, OFAC sanctions, LLM firewall, GuardGate, GuardScore® API, Kill Switch) and fan out to a cryptographically signed verdict. Pay per call in USDC on Base.

AGENTGUARD™
Architecture
Listed on x402.org/ecosystem · Sub-50ms · Cryptographically signed
02 · AgentGuard™. Agent Compliance Layer
AGENTGUARD™
one api call · cryptographically signed verdict · by MerchantGuard™
01
KYA
/
Know Your Agent · cross-platform identity lookup
<10ms
02
OFAC SDN
/
sanctions screening · live Treasury XML
<5ms
03
LLM Firewall
/
Sentinela L0→L1→L2→L3 · address detection
<10ms
04
GuardGate
/
capability-gated trust verification
<2ms
05
GuardScore™ API
/
4-component composite risk score
<20ms
06
Kill Switch
/
FORCE_3DS · BLOCK · FREEZE actuator
<2ms
Trust Attestation
multi-agent provenance · Ed25519-signed audit chain
USDC payment on BaseStripe charge / disputeConversational messageAI agent tool callCross-platform agent identity
✓ALLOW✕DENY!Enhanced Screening◇Honeypot
AgentGuard™ USPTO 99462472 published Apr 14·4 patent provisionals filed Feb 17, 2026·Listed live on x402.org/ecosystem·Sub-50ms total per call
MerchantGuard™ · The compliance layer for the AI agent economy
merchantguard.ai · @merchantguardhq
See architecture →Pay-per-call · USDC →

The compliance stack.

Drop-in primitives for agent identity, conversational fraud, and active defense. Plug into any stack.

PAID$299–$999/mo

Sentinela API

Conversational intent firewall. Screen every agent message before it responds. 161 patterns × 6 languages. Sub-10ms.

PAIDFrom $2,500/mo

Kill Switch

Active VAMP defense for ISOs and PSPs. Auto-fires FORCE_3DS at 1.3% with graduated response. HMAC-signed webhooks.

PAIDContact sales

GuardGate

Agent authorization firewall. Capability-gated trust verification. Block agents below threshold from sensitive actions.

FREE OSSMIT licensed

AgentGuard CB

Open-source Stripe dispute evidence compiler. CE 3.0 packets, signed audit trail, no LLM. Free on npm + GitHub.

USPTO PATENT-PENDING

Powered by 4 patent-pending primitives.

Every product traces back to a filed USPTO provisional. The IP backbone for the agent compliance category.

A63/983,615

Federated Trust Sharing

Privacy-preserving cross-platform agent reputation via OPRF and zero-knowledge proofs.

B63/983,621

Phonetic Cascade

L0–L3 conversational fraud detection with reverse phonetic folding for cross-lingual coverage.

C63/983,843

KYA Trust Oracles

Know-your-agent verification, LLM API firewall, and Kill Switch active defense in one stack.

D63/984,626

DAG Trust Attestation

Cryptographic chain of custody for multi-agent transactions. Production-deployed at 34 sites.

See the patent portfolio →

Install AgentGuard™. Get protected.

One install. MCP, npm, or API. Pay per call in USDC on Base or subscribe in fiat.

Get an API key

No credit card required for the API.