MCP
Native tools for Claude, GPT-4, and any LLM that supports the Model Context Protocol
A2A
Google’s Agent-to-Agent protocol — structured capability discovery and invocation
x402
Pay per call with USDC. No subscription, no API key, no human in the loop
All discovery endpoints (
/.well-known/*, /llms.txt, /llms-full.txt) are free and unauthenticated. An agent can discover LimitGuard’s full capabilities without a wallet or API key.Why Trust Intelligence Matters for Agents
An autonomous agent executing a payment workflow faces a fundamental asymmetry: it can verify its own instructions perfectly, but it has no ground truth about the entity receiving the payment. A supplier claiming to be “Acme Corp BV” in Amsterdam could be:- A legitimate, 12-year-old company with verified KVK registration
- A recently-incorporated shell with no trading history
- An entity on an OFAC or EU sanctions list
- A domain registered last week pointing to a known fraud cluster
proceed / review / enhanced_due_diligence / block), and the evidence behind it.
Agent Decision Framework
| Trust Score | Recommendation | Suggested Agent Behavior |
|---|---|---|
| 80-100 | proceed | Execute autonomously |
| 60-79 | review | Execute with audit log entry |
| 40-59 | enhanced_due_diligence | Pause and request human confirmation |
| 0-39 | block | Abort transaction, alert operator |
Protocol Overview
| Protocol | Discovery Endpoint | Best For |
|---|---|---|
| MCP | /.well-known/mcp.json | LLM agents (Claude, GPT-4, Gemini) |
| A2A | /.well-known/agent.json | Agent-to-agent workflows (Google ADK, LangGraph) |
| x402 | /.well-known/x402.json | Autonomous agents paying per call with USDC |
| llms.txt | /llms.txt, /llms-full.txt | Context loading for any LLM session |
MCP (Model Context Protocol)
MCP is Anthropic’s standard for exposing APIs as native tools to LLM agents. When an LLM is configured with LimitGuard’s MCP manifest, it can call trust verification endpoints the same way it calls any built-in tool — no prompt engineering required.Discovery
Response
Configuring an LLM Agent
A2A (Agent-to-Agent)
Google’s Agent-to-Agent protocol defines a standard for agents to discover and invoke other agents as structured services. LimitGuard publishes an Agent Card that exposes its capabilities, skills, and invocation interface to any A2A-compatible agent.Discovery
Response
How an Agent Uses A2A
x402 Micropayments
x402 is the HTTP native payment protocol for AI agents. An agent with a funded USDC wallet can call LimitGuard endpoints without a subscription or API key — it discovers pricing, builds a cryptographic payment signature, and includes it in the request header. This is the fully autonomous path: no human creates an account, no API key is provisioned, no billing is configured.Discovery
Response
How x402 Works
Probe the endpoint (optional)
Make a request without payment. The server returns HTTP 402 with the exact payment requirements for this call. You can skip this step if you already read requirements from
/.well-known/x402.json.Build the payment signature
Construct an EIP-3009
TransferWithAuthorization signature authorizing the USDC transfer. Sign it with your agent’s wallet private key.Base64-encode the payment object
Serialize the payment payload to JSON and base64-encode it. This becomes the
X-PAYMENT header value.Retry the request with X-PAYMENT
Send the original request again, including the
X-PAYMENT header. LimitGuard verifies the signature on-chain, executes the transfer, and returns the trust score.Full x402 Implementation
Quality Tiers
Control cost vs. freshness with theX-Response-Quality header. The payment amount must match or exceed the tier price.
| Tier | Header Value | /v1/entity/check | /v1/risk/score | /v1/kyb/check | Data Freshness |
|---|---|---|---|---|---|
| Cached | cached | $0.01 (10,000) | $0.005 (5,000) | $0.25 (250,000) | Up to 24h cached |
| Fresh | fresh | $0.05 (50,000) | $0.01 (10,000) | $1.25 (1,250,000) | Live fan-out (default) |
| Enhanced | enhanced | $0.15 (150,000) | $0.05 (50,000) | $2.50 (2,500,000) | Live + premium sources |
End-to-End Agent Example
This example shows a complete autonomous payment agent — from cold start to decision — using discovery, x402 payment, and trust-gated execution.Complete Implementation
llms.txt — LLM Context Loading
LimitGuard implements the llms.txt standard — machine-readable files that let any LLM load a structured API summary into its context window.Endpoints
| File | Size | Use |
|---|---|---|
/llms.txt | ~2KB | Quick context loading — overview + key endpoints |
/llms-full.txt | ~20KB | Complete reference — all endpoints, pricing, examples |
When to Use llms.txt
Cross-Sell Recommendations
Agents calling one endpoint can discover complementary services at runtime:Response
Sandbox Mode for Agent Development
Build and test your agent integration without spending USDC or calling real data sources.Activating Sandbox for Agents
Two approaches depending on your agent’s authentication method:Sandbox Behavior Reference
| Property | Sandbox | Production |
|---|---|---|
| Data sources | Mock only — no real KVK, sanctions, VIES | All 8 sources, live |
| Responses | Deterministic (same input = same output) | Live results |
| x402 payment | Skipped entirely | Required (or API key) |
| Quota | Not consumed | Consumed |
| Rate limit | 10 req/min per IP | Tier-based |
| Response time | ~5ms | ~250-500ms |
Writing Agent Tests with Sandbox
Best Practices
1. Cache Discovery Responses
Discovery endpoints don’t change between deploys. Cache them aggressively.Discovery endpoints have a
Cache-Control: max-age=3600 header. Respect it — refreshing more frequently adds latency with no benefit.2. Handle HTTP 402 Gracefully
A well-built agent treats 402 as a normal flow, not an error.3. Use Quality Tiers Strategically
Don’t pay forfresh when cached is sufficient.
| Scenario | Recommended Tier | Reason |
|---|---|---|
| First-time supplier onboarding | enhanced | High stakes, need maximum signal |
| Repeat invoice from known supplier | cached | Already verified, use cached result |
| Pre-screening inbound leads | cached | Volume operation, cost matters |
| Transaction >$10,000 | fresh | High value, needs live data |
| Compliance audit trigger | enhanced | Full evidence trail required |
4. Log Trust Scores in Your Audit Trail
For regulated workflows, store the trust score alongside the transaction:5. Set Nonce Expiry Windows Conservatively
x402 nonces expire 5 minutes aftervalidBefore. Set a 5-minute window (validBefore = now + 300) — tight enough to prevent replay attacks, wide enough to survive network retries.
6. Never Hardcode Facilitator Addresses
The facilitator address is returned dynamically in each 402 response. Always read it from the response — never hardcode it. The address can change during upgrades.Protocol Comparison
| MCP | A2A | x402 | |
|---|---|---|---|
| Best for | LLM agents (Claude, GPT) | Multi-agent workflows | Fully autonomous agents |
| Authentication | API key via manifest config | API key or x402 | USDC wallet (no account) |
| Discovery | /.well-known/mcp.json | /.well-known/agent.json | /.well-known/x402.json |
| Human setup required | Minimal (configure MCP server) | Minimal (discover card) | None |
| Cost model | Subscription or quota | Subscription or quota | Pay-per-call USDC |
| Capability negotiation | Tool definitions | Skill catalog | Endpoint + pricing listing |
| Production maturity | GA | Beta | GA |