What 3,668 Agent API Calls Tell Us About the Agent Economy

August 2, 2026 · Data from minia2a.uk trial network · 316 developers, 92 endpoints
3,668
Total Trial API Calls
316
Unique Users
92
Endpoints with Trials
34
Wallet Users (Paying)

The Top 10: What Agents Actually Call

Ranked by trial volume. The "Unique Users" column shows how many different developers are building in each category — a proxy for developer intent:

#EndpointTrialsUnique UsersCategory
1x402-captcha-solve398Anti-bot
2x402-gas358Blockchain
3x402-web-scrape25725Web Tools
4x402-find226Knowledge
5x402-store187Storage
6x402-recall158Memory
7x402-token-security11319Security
8x402-polymarket109Prediction Markets
9x402-time10815Utility
10x402-swap-safety721DeFi Security

Three Patterns That Jump Out

1. Agents Need Infrastructure, Not AI

The top 3 endpoints — captcha solving, gas estimation, web scraping — aren't "AI" services. They're infrastructure that agents call to interact with the world. An agent browsing the web hits captchas. An agent transacting on-chain needs gas prices. An agent researching needs to scrape pages.

This matches a broader truth: agents are software that needs the same plumbing humans do — just programmatically. The highest-volume services are bridges between agents and human-oriented infrastructure.

💡 What This Means for Builders

Don't build another "AI summarizer" API. Build the pipes agents need: reliable web scraping with automatic captcha solving, chain-abstracted gas estimation, persistent key-value storage with TTLs. The demand is in the boring infrastructure, not the flashy AI.

2. The Long Tail Is Long — 92 Endpoints With Real Usage

Beyond the top 10, another 82 endpoints have meaningful trial volume. Everything from DNS lookups (20 trials) and IP geolocation (22) to token search (16), QR code generation (16), and sentiment analysis (33).

Usage by category:

CategoryKey EndpointsEst. Combined Trials
Web & Dataweb-scrape, web-retrieve, screenshot, text-scrape, http-analyze, fetch~500
Blockchain & Cryptogas, token-security, swap-safety, polymarket, wallet-intel, domain-intel, dex-price, funding-rate, onchain-reads, chain-registry, abi-lookup, tx-decode, wallet, ens-resolve, address-parse~900
Knowledge & Searchfind, polymarket-scan, search, keywords, defi-directory, bounties, ecosystem-health, trading-signal, fear-greed, agent-skill-pack, prompt-templates~550
Storage & Memorystore, recall, data-enrich~360
Utilitycaptcha-solve, time, uuid, base64, hash, jwt-decode, math, random, qr, csv-json, text-to-json, url-codec, text-diff, text-stats, regex-explain, language-detect, markdown, emoji, color, country, date-math, dns, ip-lookup, ping, uptime~1,080
Security & Validationemail-verify, liveness-check-v1, account-age, guardian, npm-audit, pip-audit~145
AI & Contentsummarize, sentiment, api-review, agent-toolkit, pdf-text, render~140

Blockchain/crypto is the largest category by both endpoint count and volume. This makes sense: agents deploying on-chain need reliable, structured data about gas, tokens, security, prices, and transactions — things that are hard to scrape from block explorers but easy to expose as x402 APIs.

3. The Conversion Funnel: 316 → 34 → 14

The developer journey tells its own story:

StageCountConversion Rate
Trial users (made ≥1 free call)316
Wallet users (registered USDC wallet)3410.8%
Paid transactions (real USDC)1441% of wallet users

A 10.8% trial-to-wallet conversion is healthy for developer infrastructure. But the 14 paid transactions ($12.75 total volume) show we're still early: developers are testing, not yet running production workloads. The 34 wallet users represent real intent — they connected a wallet because they plan to pay.

🔑 The Wallet Gap

The biggest drop-off isn't from trial to wallet (10.8% is solid). It's from wallet to regular usage. Why? Because most agent frameworks don't yet have built-in x402 payment support. Developers who connect a wallet still have to write custom payment logic. This gap closes as SDK support grows (wayleave, centinel, level42, agentwallet-sdk).

What 316 Developers Are Building

The user distribution by endpoint reveals clusters of developer intent:

The Bottom Line

After 3,668 API calls across 92 endpoints from 316 developers:

  1. Infrastructure beats AI. Agents pay for pipes — captcha solving, gas data, web scraping — not for "smart" APIs.
  2. Crypto-native agents are the early adopters. Blockchain/crypto endpoints dominate. These developers already understand programmable money, making x402 a natural fit.
  3. The market is real but early. 316 developers trying, 34 connecting wallets, 14 paying. The demand is there; the tooling is catching up.
  4. Long tail = opportunity. 92 endpoints with real usage means agent developers want everything programmatically accessible with micropayments attached. Every API is a candidate for x402.

🔗 Explore the full marketplace → · 📊 Live stats → · 🛠️ Get 500 free credits →