DEV Community: Karim Gueye

What AI agents actually pay for — six weeks of data from 101 pay-per-call endpoints

Karim Gueye — Wed, 29 Jul 2026 18:09:03 +0000

A few weeks ago I wrote up what agents were paying for on NetIntel, my platform of pay-per-call APIs settled in USDC over x402 — no signup, no API keys, no accounts. An agent hits an endpoint, gets a 402 Payment Required, pays a fraction of a cent, and gets structured data back. That's the whole loop.

Since then the dataset has grown, I've instrumented every settled call into a proper database (payer wallet, endpoint, price, latency, transaction hash), and I've launched a second settlement rail. So this is the rewrite with real numbers instead of eyeballed ones — and the findings didn't soften. They sharpened.

The setup

2,646 settled paid calls from 194 distinct paying wallets, across 101 live endpoints, over six weeks of instrumented production data. Every call in this dataset is a real on-chain payment with a transaction hash — no test traffic, no estimates. Settlement runs on Base, and as of this month on Solana too.

This is still one platform's data in a young ecosystem — the caveats are at the bottom, and one of them is bigger than it looks. But the shape has now held for six weeks straight, and it's the same shape I flagged the first time.

Finding 1: revenue is absurdly concentrated — and it stayed that way

Five endpoints drive 69% of all revenue. One of them — a text-to-structure endpoint that takes messy input and returns strict typed JSON — is 42% by itself. The rest of the top five are all in the same family: translation, structured LLM inference, and one domain-intelligence report.

The other 96 endpoints split the remaining 31%. Thirty-five of the 101 have never been paid for once. Not "underperformed" — zero settled calls, ever.

When I first published this pattern I wondered if it was an artifact of a small sample. The dataset has since more than doubled and the concentration ratio barely moved. I now treat it as the market talking, not noise.

Here's the part I'd want to know if I were reading this: that 42% endpoint is essentially one buyer — a wallet running a daily cron that has settled a payment nearly every day for six weeks. Mid-month I doubled that endpoint's price. The buyer didn't pause for a single day. When an agent has wired your endpoint into a workflow that works, the price of the call is nothing next to the cost of replacing it. Which leads directly to —

Finding 2: agents pay to fix text, not for exotic data

I built NetIntel as a network intelligence platform — DNS forensics, IP reputation, certificate transparency, the works. What actually earns is transformation: schema extraction, translation, structured inference, sentiment. Text goes in messy, structure comes out guaranteed.

Four of the top five earners are transformation endpoints. The network-intelligence catalog I was proudest of — the technically hardest stuff — makes up most of the 35 endpoints nobody has ever paid for.

The reframe from the first article still holds, so I'll restate it plainly: an agent's expensive problem isn't accessing data, it's trusting the shape of what it already has. A blob of text it must parse itself costs tokens, a fragile parsing step, and an entire error branch for "what if this comes back malformed." A guaranteed clean structure removes that liability in one call. The most valuable thing I sell isn't information — it's the removal of the agent's own uncertainty.

Finding 3: volume is not value

This one got more extreme with more data. The two most-called endpoints on the platform — a cheap utility lookup and a content extractor, together nearly 35% of all call volume — combine for 6.3% of revenue. The third-most-called endpoint, on nearly identical call volume, is the text-to-structure endpoint making 42%.

Spend tracks the value of the task completed, not the frequency of the call. If you optimize for the metric that's easiest to see on a dashboard — call count — you will double down on exactly the wrong endpoints.

Finding 4 (new): cheap endpoints are the front door

This is the one the new instrumentation surfaced, and it complicates the "kill the breadth" conclusion from my first write-up.

I track which endpoint each wallet paid for first. The winner isn't close: my cheapest utility endpoint was the first paid touch for 39% of all wallets. And those wallets went on to spend, across the whole catalog, 3.4× what that endpoint itself has ever earned.

The pattern repeats down the acquisition table: cheap, low-stakes, easy-to-verify calls are how agents try a new provider. Then the cross-sell data shows them fanning out — more than half of all wallets have paid for at least two different endpoints, and 92% of wallets came back for more than one paid call.

So the catalog isn't just dead weight plus five winners. It's a funnel:

Cheap utility endpoints acquire. An agent risks one cent to learn you return clean data and settle payments correctly.
Transformation endpoints monetize. Once trust is established, the high-value calls follow.

The 35 never-called endpoints are still dead weight. But I'm no longer treating "low revenue" alone as a kill signal — the question for each endpoint is now does it earn, or does it acquire? An endpoint that does neither, goes.

What I changed

Depth over breadth, still. I stopped shipping speculative endpoints to "see if they stick." I know what sticking looks like now, and most never will. Investment goes into making the winners faster, more reliable, and harder to leave.
Priced by value, not by cost. The price-doubling experiment on the top endpoint was the direct result of finding 3. Revenue per call doubled; volume didn't move.
Kept the cheap funnel open. Findings 1–3 alone say "trim the menu." Finding 4 says the cheapest endpoints are marketing that pays me. They stay.
Added a second rail. Solana settlement went live across every endpoint this month, next to Base — same catalog, same prices, agent's choice. It's already 7% of settled calls.

Caveats (read these before you quote me)

Small sample. ~2,600 paid calls is enough to see a shape, not to bank a law. Directional signal from one operator, not a benchmark.
Wallets ≠ customers. This is the caveat I underweighted last time. When I clustered wallets by their funding source, a large share of my "distinct wallets" traced back to a single operator running many wallets — and when that operator left mid-month, my new-wallet numbers dropped with them. Anyone quoting "unique wallets" in this ecosystem (including me, above) is overcounting customers. Cluster by funder before you believe your own dashboard.
The 42% is one buyer. A loyal, price-insensitive, six-weeks-daily buyer — but one buyer. Concentration cuts both ways: it's my strongest product signal and my biggest single point of revenue failure.
One platform, one catalog. My menu skews network-intel and text tools. Your mix would surface different winners. The concentration pattern may generalize; the specific winner probably won't.
Early ecosystem. A meaningful share of x402 traffic is still experimental. I've leaned on paid, repeat behavior (the 92% repeat rate helps), but the noise floor is real.

None of that changes the headline, which has now survived a doubling of the dataset: agent spend is extreme in its concentration, it favors transformation over raw data, it ignores call volume entirely — and the cheap calls that earn nothing are how the expensive relationships start. If you're building for this market, plan for a few things carrying everything, figure out which few as fast as you can, and don't delete the front door while you're trimming.

I run NetIntel — pay-per-call network and structured-data intelligence for agents over x402, on Base and Solana. If you're building agents and want to compare notes on what your traffic actually pays for, I'm genuinely interested; that's the data I wish more people published.

What AI agents actually pay for — data from 74 pay-per-call endpoints

Karim Gueye — Thu, 09 Jul 2026 13:05:37 +0000

For the last several months I've run NetIntel, a platform of pay-per-call APIs settled in USDC over x402 — no signup, no API keys, no accounts. An agent hits an endpoint, gets a 402 Payment Required, pays a fraction of a cent, and gets structured data back. That's the whole loop.

I built a lot of endpoints. Network intelligence, domain forensics, text transforms, a long menu of things I was convinced agents would want.

Then I looked at what they actually paid for. The answer wasn't what I expected, and it changed how I build. Here's the data, with the caveats up front so nobody has to dig for them.

The setup

Roughly 2,000 paid calls (per x402scan, the public x402 analytics tracker) across 74 live endpoints, real USDC on Base, over a period of months. This is one platform's data in a young ecosystem — see the caveats at the bottom before you extrapolate. But the shape of it was clear enough that I stopped treating it as noise.

Finding 1: revenue is absurdly concentrated

Five endpoints drove 75% of all revenue. One of them — a text-to-structure endpoint that takes messy input and returns strict typed JSON — was 40% by itself.

The remaining ~69 endpoints split what was left, and most of them earned close to nothing. Not "underperformed" — nothing. I had spent weeks building endpoints that, in production, no agent ever paid for twice.

If you've built for agents, sit with that ratio. I had assumed a broad menu was an asset: more surface area, more ways to get discovered, more shots on goal. In practice the breadth was dead weight. The menu didn't compound; five items carried it and the rest were maintenance cost.

Finding 2: agents pay to fix text, not for exotic data

The thing that dominated wasn't clever data. It was transformation — turning unstructured text into a schema the agent could act on.

That reframed the whole product for me. An agent's expensive problem isn't accessing data; it's trusting the shape of what it already has. A blob of text it has to parse itself is a liability — it costs tokens, it costs a fragile parsing step, it costs an entire branch of "what if this comes back malformed." Selling it a guaranteed clean structure removes that liability in one call.

The most valuable thing I sell isn't information. It's the removal of the agent's own uncertainty. That's a different business than "API for X data," and I only saw it by looking at the receipts.

Finding 3: volume is not value

The endpoint agents called most often was not the endpoint that made the most money. Not close.

Spend tracked the value of the task completed, not the frequency of the call. High-frequency, low-stakes lookups generated a rounding error. Lower-frequency, high-value transformations generated the revenue. If I'd optimized for call volume — the metric that's easiest to see on a dashboard — I'd have doubled down on exactly the wrong endpoints.

What I changed

I killed the breadth instinct. I stopped shipping speculative endpoints to "see if they stick," because now I know what sticking looks like and most of them never will. The plan is depth on the few categories that actually pay: make the winners faster, cheaper, more reliable, and harder to leave, rather than adding a 75th thing nobody asked for.

Breadth felt like progress because building is satisfying and each new endpoint looked like an asset on the menu. The data says it was mostly a distraction I could afford to build only because it was cheap to build. That's not a strategy — it's a hobby that happens to be adjacent to a business.

Caveats (read these before you quote me)

Small sample. ~2,000 paid calls is enough to see a shape, not enough to bank a law. Treat this as a directional signal from one operator, not a benchmark.
One platform, one catalog. My menu is skewed toward network/domain intelligence and text tools. Your mix would surface different winners. The concentration pattern may generalize; the specific winner probably won't.
Early ecosystem. A meaningful share of x402 traffic today is experimental and non-commercial. Some of what looks like demand is people kicking tires. I've tried to reason about paid, repeat behavior rather than one-off test calls, but the noise floor is real.
Survivorship in the framing. "The other 69 were a distraction" is the lesson I drew, but a few of those may just be early, or badly discovered, rather than genuinely unwanted. I'm betting they're dead weight; I could be wrong about the margins. None of that changes the headline: in production, agent spend was extreme in its concentration, it favored transformation over raw data, and it ignored call volume entirely. If you're building for this market, I'd plan for a few things carrying everything — and figure out which few as fast as you can.

I run NetIntel — pay-per-call network and structured-data intelligence for agents over x402. If you're building agents and want to compare notes on what your traffic actually pays for, I'm genuinely interested; that's the data I wish more people published.

Built NetIntel: 64 network-intelligence + data tools for AI agents

Karim Gueye — Sun, 21 Jun 2026 23:45:16 +0000

Built NetIntel: 64 network-intelligence + data tools for AI agents.

DNS, SSL, WHOIS, email security, OSINT, structured data extraction — all pay-per-call via x402 on Base. No API keys, no signup. Agents just pay a fraction of a cent and go.

Live now 👇
github.com/kjgueye/netintel-api

@CoinbaseDev @x402