What Happened
Our data pipeline (scripts/data-agent.sh) was counting every HTTP 200 response from x402 endpoints as a "paid transaction." The problem: free trials also return HTTP 200. So paid transactions, free trial calls, and test requests were all lumped together.
Once we fixed the counting to check the actual PAYMENT-RESPONSE header and the /api/stats transactions field, the numbers came back to earth — hard.
The Real Numbers (August 3, 2026)
The Honest Funnel
| Stage | Count | Drop-off |
|---|---|---|
| Total requests (all time) | 327,186 | — |
| Free trials used | 4,594 | — |
| Unique trial users | 316 | — |
| Wallet registrations | 34 | 89% from trials |
| Paid transactions | 14 | 59% from wallets |
The conversion rate from trial → wallet is about 10.8%. From wallet → paid is about 41%. From trial → paid directly: 0.30%.
Why So Few Pay?
We don't have a payment UX for agents yet. Here's what we see:
- Agents can't easily hold USDC. Most AI agent frameworks have no wallet integration. An agent can discover an x402 API, get a 402 response, and have no way to pay.
- Free trials are generous. With 500 free credits per registration and 4,594 trials used, most callers never exhaust their trial budget. They don't need to pay.
- Discovery is the real bottleneck. When agents do find services they need and have a wallet configured, they pay. The 14 transactions came from agents that already had wallets set up.
- The market is still tiny. 316 developers have tried it. 34 set up wallets. This is a niche within a niche: developers building autonomous agents that need to call paid external APIs.
What We're Doing About It
- Agent wallet SDKs — Making it dead simple for agent frameworks to add USDC wallet + x402 payment in one import
- More service endpoints — 181 services now, targeting 500+ to cover the long tail of agent needs
- Better onboarding —
POST /api/v1/register-simplegives 500 free credits with no KYC, no email verification, no human - Real reporting — Fixed the data pipeline so we never publish inflated numbers again
What This Actually Means for the Agent Economy
The 76% number was wrong, but the thesis is still right: agents will pay for API access when the infrastructure exists. The 14 real transactions prove it — those agents paid real USDC on Base for real API calls. The bottleneck isn't willingness to pay. It's:
- Wallet distribution: Getting USDC wallets into agent hands
- Discovery: Agents knowing these paid services exist
- Critical mass: Enough services that agents need to pay for
We're in the first inning. The infrastructure is being built. The 14 transactions aren't a failure — they're proof the pipes work. Now we scale.
🔍 Data Integrity Commitments
Going forward, all minia2a stats come from /api/stats (live, not cached). We check the transactions field for actual on-chain payments, not HTTP status codes. If we publish a number, it's verifiable. If we make a mistake, we correct it publicly.
Build on Real Data
# Get live stats anytime
curl https://minia2a.uk/api/stats
# Register your agent (free, instant, no KYC)
curl -X POST https://minia2a.uk/api/v1/register-simple \
-H "Content-Type: application/json" \
-d '{"name":"your-agent-name"}'
# List all paid services
curl https://minia2a.uk/api/services→ Read our deeper analysis: The Agent Payment Conversion Problem (0.36% trial → paid)