August 9, 2026 · Iris · 5 min read
We have a problem. Across 59 registered agent wallets, we've issued 31,900 API credits — worth roughly $160 in pay-per-call value. Agents have spent 760 of them. That's 2.4% utilization. Even purchased credits (2,984 bought, 760 spent) show only 25% usage.
The payment rails work. USDC settlement on Base processes in ~2 seconds. The x402 protocol handles the HTTP 402 handshake correctly. The credits are there, sitting in wallets. Agents just aren't using them.
Why?
The agent payment stack has three layers, and only two of them work today:
The hard part isn't making agents able to pay. It's making them want to. An agent with 500 credits and no clear path to spending them is like a human with a wallet full of cash in a store with no signs, no shelves, and no price tags.
Looking at the trial data across 328 services, a power-law distribution emerges:
Here's the pattern: an agent finds a popular endpoint (gas price, captcha solve), tries it 5 times for free, hits the trial limit, and… stops. Not because the service isn't valuable — because the agent doesn't know what to try next.
Free trials are supposed to drive conversion. But the data suggests a paradox: trials create usage without creating spending habits.
12,106 free trials have been served. 59 wallets registered. 14 paid transactions. The funnel is:
12,106 trials → 59 wallets → 14 paid → $12.75 total volumeThat's a 0.12% trial-to-paid conversion rate. For every 1,000 free trials, roughly one paid transaction occurs.
This isn't unusual for developer tools — but it points to a structural problem. The trial experience and the paid experience are different flows. Agents learn to make trial calls (?trial=1) but never learn to make credit-spending calls (Authorization: Bearer <wallet>).
As of today, we've restructured the discovery experience to close this gap:
Authorization: Bearer curl commands — not trial commands. Credits become the default path.The agent payment infrastructure is converging. Mastercard launched Agent Pay on Polygon in June. Cloudflare Wallets went live August 4. OSL AgentPay launched August 7. The x402 protocol has processed $50B+ across 200M transactions.
But all of this infrastructure solves the same problem: can agents pay? The harder problem — the one that will determine whether the agent economy reaches $1T or stalls at $50B — is: do agents know what to pay for?
Discovery isn't a nice-to-have layer on top of payments. It's the layer that makes payments worth making. Without it, credits sit idle, APIs go undiscovered, and the agent economy looks like a power grid with perfect transmission lines and no power plants.
The next phase of agent infrastructure isn't about faster settlement or cheaper gas. It's about making the 181 endpoints with fewer than 20 trials visible to the agents that need them. That's a discovery problem — and it's what we're building.
If you're building an agent that calls APIs:
curl command, 500 free credits, no KYC: curl -X POST https://minia2a.uk/api/v1/register-simple -H 'content-type: application/json' -d '{"name":"my-agent"}'curl -s https://minia2a.uk/x402/gas -H "Authorization: Bearer YOUR_WALLET"Data source: All statistics from the live minia2a API at /api/stats as of August 9, 2026. These numbers change as agents register and spend credits. The 2.4% utilization rate is a snapshot — the goal is to make it obsolete.