Top Endpoints Command $18,000 in Monthly Payments for AI Agents

The most active x402 endpoints on Lumière PayCheck are generating significant revenue, with the single highest-performing API processing nearly $18,000 in payments over a 30-day period. This concentration of value among a small group of verified providers highlights where reliable, high-volume payment infrastructure currently exists for AI agents.

The Concentration of Value in Verified APIs

When building agents that pay for APIs, it is crucial to understand where the money is flowing. Our data shows that payment volume is not evenly distributed across all watched endpoints. Instead, a small number of providers handle the vast majority of transactions.

The top endpoint by real payment volume over the last 30 days is ax1-new.vercel.app. This single API processed $17,755 in USDC on Base, handling 887,840 payments from 8,008 unique buyers. The median cost per call was just $0.02, yet it maintained a perfect Grade A score of 100/100 and 100% uptime.

Other endpoints also show strong performance but at lower volumes. For instance, arbipulse.theaslangroupllc.com and mailpulse.theaslangroupllc.com each processed $219.14 in payments from 59 buyers, with a median cost of $0.10 per call. Similarly, astro.openverbs.com handled $90.75 in payments from 54 buyers at a median cost of $0.04 per call.

For agent builders, this data suggests that relying on endpoints with established payment history reduces the risk of failed transactions. While many endpoints are watched and graded, only those with verified real payment delivery have proven they can consistently handle the financial mechanics required for automated billing. The top 100 endpoints by score all have delivery verified by a real paid test, providing a safer starting point for integration.

Understanding the "Caution" Verdict

A common misconception is that an endpoint requiring caution is unsafe or broken. In reality, the "caution" verdict primarily indicates that we have not yet confirmed delivery through a paid test. Out of 119,050 endpoints watched as of 2026-11-03, 74,960 (63.0%) fall into this category.

These endpoints may cost more per call than our tester pays for, or they might require user sign-in before providing a valid payment quote. They are not necessarily risky; they simply lack the confirmed proof of delivery that "proceed" endpoints have. Only 21,165 endpoints (17.8%) currently recommend proceeding without hesitation, and even among these, only some have been paid-tested.

The "avoid" category contains 5,279 endpoints (4.4%), which failed our checks due to being down, having broken payment quotes, or failing a paid test delivery. This is a small fraction of the total landscape, but it underscores the importance of filtering before integrating. Builders should view the "caution" list as a pool of potential candidates that need further verification rather than a list of dangerous sites.

Reliability in Monitoring and Settlements

The accuracy of these grades depends on rigorous monitoring. In our latest monitoring cycle, we performed 43,565 checks in just 35 minutes. Of these, 42,627 were answered correctly, resulting in a 97.8% success rate for the grading process itself.

We do not rely on simulated payments alone. Lumière PayCheck reads real settlements on Base to verify that funds actually reached the endpoint’s payout wallet. This ensures that the "Grade A" status is backed by financial reality, not just technical availability.

For sellers, this means that visibility in the top tiers of our list correlates with actual revenue generation. The endpoints listed above are not just technically sound; they are financially active. For buyers, it means that the grading system reflects real-world performance, helping agents avoid endpoints that might appear functional but fail to process payments correctly.

How We Measured

PayCheck checks every endpoint listed in the x402 Bazaar on a schedule, makes small real paid test calls, and reads real settlements on Base. This combination of technical verification and financial confirmation provides a more accurate picture of an endpoint’s readiness for AI agent integration than standard uptime monitors.

What to Do

* Check before paying: Use the Lumière PayCheck API to verify the current grade of any endpoint your agent intends to call. Look for "proceed" verdicts and confirmed paid delivery. * Set spend limits: Even with verified endpoints, costs can vary. Set strict budget limits in your agent’s code to prevent unexpected expenses from high-volume calls. * Prefer verified endpoints: Prioritize endpoints that have been graded A and show real payment volume. This reduces the likelihood of integration errors and ensures smoother billing processes for your users.

The full live list is at https://lumierepaycheck.org/stats and the API is free at https://lumierepaycheck.org/v1/stats