[Feature Request] Add spending controls / circuit breakers for agent payments

[mkmkkkkk]

Hey @microchipgnu — love what you're building with MCPay.

One gap I keep running into with x402 MCP payment flows: there's no guardrail layer between an agent deciding to pay and the payment executing. In production, this means:

• Agent retries a failed request 50 times → 50 charges
• Prompt injection tricks agent into paying a drainer address
• Daily spend blows past budget with no alert
• No way to enforce per-endpoint spending policies across MCP servers

I'm building PaySentry — a control plane that sits between the agent and the payment protocol. It adds:

• Per-agent spending limits (daily/hourly/per-tx)
• Endpoint whitelisting
• Circuit breakers on anomalous patterns
• Full audit trail

The @paysentry/x402 adapter wraps any x402 facilitator client with policy enforcement:

import { PaySentryX402Adapter } from '@paysentry/x402';

const adapter = new PaySentryX402Adapter(
  { policyEngine, spendTracker },
  { circuitBreaker: { failureThreshold: 5, recoveryTimeoutMs: 30_000 } }
);

adapter.withLifecycleHooks(yourX402Server);

Integration Points

I see a few natural touch points with MCPay:

1. Monetizer Proxy Layer: Add optional PaySentry middleware before payment execution
2. Client SDK: Wrap the wallet adapter with spending policy enforcement
3. Registry: Flag MCP servers that support PaySentry-compatible policies

This would address common pain points I've seen:

• Agents running wild on Base/Avalanche with no spending cap
• No way to block suspicious payment patterns in real-time
• Per-token pricing blowing up when agents loop on bad responses

Would love to build an integration with MCPay. Happy to do the work and submit a PR. Would that be welcome?

---

Context:

• npm: @paysentry/x402 (v1.0.0)
• 79 tests passing, full E2E example
• Addresses x402 issues #1062, #808, #803