Every marketplace operator will eventually face this problem. It's not technical — the payment rails work, credits are issued, APIs are live. It's behavioral: agents sign up, receive 500 free credits, and never spend them.
Here is the raw funnel data as of August 9, 2026, from the live stats endpoint:
18.7% trial-to-registration conversion is decent for a developer tool. But the drop from 60 wallets to 788 credits spent is the activation gap:
After tracing the full user journey — landing page → discover → trial → registration → credit spending — four factors emerge:
When a developer visits the discover page, the quick-start shows:
curl -s "https://minia2a.uk/x402/gas?trial=1"
This is frictionless: copy, paste, get data. No registration. The credit-spending path — pasting a wallet address to get personalized curl commands with -H "x-wallet: 0x..." — requires the user to (a) register, (b) save their wallet address, (c) return to the discover page, (d) paste it. That's four steps after the initial "aha" moment. Most users never make it past step (a).
Registration takes 10 seconds. No KYC. No gas. That's the strength — and the weakness. Users register because it's easy, not because they have an immediate use case. The 500 free credits feel like "free money" to claim now and use later. "Later" never arrives.
This is consistent with the broader x402 ecosystem: 82,718 services, 21.9M transactions, only 381 identified AI agent buyers (1.3%). The activation gap isn't a minia2a problem. It's a category problem.
After registration, the user receives a wallet address and private key. Then: silence. No "you still have 500 credits" reminder. No weekly summary of what endpoints have new features. The credits sit dormant because nothing nudges the user back.
Human SaaS products have onboarding emails and "complete your profile" flows. Agent marketplaces tend to treat registration as the finish line, not the starting line. The agent is expected to autonomously discover, evaluate, and pay — but the agent doesn't even know it has credits unless the developer wires it up.
An agent developer doesn't browse a marketplace. They encounter a need in their IDE — "I need gas prices for this DeFi bot" — and choose: (a) write a free API wrapper (slow), or (b) call a paid endpoint (fast, costs credits).
The "spend credits" decision happens in the IDE, not on a marketplace website. If the developer doesn't have their wallet address memorized or the curl command handy, they choose option (a) — build it themselves. Credits only get spent when the paid path is genuinely easier than the DIY path.
| Metric | Value |
|---|---|
| x402 all-time transactions | 200M+ (Linux Foundation) |
| AI agent buyers identified | 381 of 28,000+ wallets (1.3%) |
| Top x402 seller revenue (on-chain) | $0.70 (1,062 wallets analyzed) |
| Coinbase agentic.market | 2,056 services, pay-first model |
| Circle Discovery API | 900+ services, compliance-screened |
| Live x402 endpoints (independent audit) | ~24% of listed (76% dead) |
| minia2a endpoints | 328 live, all health-probed daily |
| minia2a trial-to-wallet conversion | 18.7% |
| minia2a credit utilization | 2.2% |
The pattern: the infrastructure works, demand doesn't convert. 200M transactions is real volume — but it's nearly all signaling (402 responses), not settlement. The median x402 seller earns nothing. This isn't a protocol failure. It's an activation failure. x402 solves "how do agents pay?" What no protocol can solve is "why should they?"
Credit spending now works via -H "x-wallet: 0x..." in addition to ?wallet= query param. This matches how HTTP client libraries pass auth. The discover page auto-switches curl commands from trial mode to credit mode when a wallet is pasted.
credits_remaining now appears at the top level of every API JSON response, not buried inside _trial. Agents parsing responses can track balance without extra logic.
We proposed .agent-budget — a single file where developers declare {daily_limit_usdc: 5, max_per_call_usdc: 1}. Agent frameworks (Claude Code, Codex) read it and auto-approve payments within limits. The activation gap closes when spending credits becomes the default behavior, not a per-call decision.
When a registered wallet hasn't spent credits in 7 days, the next trial request from their IP returns their wallet address and a ready-to-copy credit-spending curl in the response. Not an email — a response header. Meet the agent where it already is.
Every x402 marketplace will hit this wall. The one that solves activation — turning free credits into habitual spending — wins the agent economy. That's not about who has the most APIs or the lowest fees. It's about who makes spending credits the path of least resistance for a developer mid-flow, in their IDE, at the exact moment they need an API.
The HTTP 402 standard works. The payment rails are live. 200 million transactions have proven the protocol. The next chapter isn't about better payments. It's about better habits.
Free trials create discovery but not habit. 321 developers tried minia2a. 60 signed up. A handful returned to spend credits. The fix isn't cheaper credits or a better wallet UI — it's embedding credit spending at the point of need (IDE, framework, agent runtime) and making it the default interaction, not a secondary path behind the anonymous trial.
Data from minia2a.uk/api/stats as of 2026-08-09T23:34 UTC. All numbers live and verifiable. RSS feed for updates. All posts →