Render Check: machine-buyable browser verification

Research question

Can a browser-verification task normally delivered as human web-support work be expressed as a machine-facing service that another system can understand, pay for, invoke and retrieve without a human completing the transaction?

Service under test

Atinamos Render Check was designed for finished public websites. A machine buyer supplied a public URL and the service ran independent desktop and mobile Chromium sessions against that target.

The observed price in the retained public record was 0.25 USDC per check. The paid route used x402 on Base.

Machine-service flow

Seller-side production observation

Evidence returned

The service recorded browser-facing evidence rather than only returning a binary success flag. The retained public description includes desktop and mobile screenshots, HTTP/browser failures, failed requests, console errors and warnings, result JSON and SHA-256 verification hashes.

That made the output suitable for another system to inspect, while still leaving interpretation to the buyer.

What the experiment supports

The retained seller-side production evidence supports that Render Check was a real paid machine service rather than a static demonstration page: a client could encounter the payment boundary, make the required payment, create a job and retrieve browser-verification output.

It also demonstrated an important design pattern used later across Atinamos: machine-readable service identity, explicit price and input, asynchronous execution, result retrieval and evidence that can be consumed by another machine.

What it does not support

This observation does not establish broad external demand, permanent service reliability, universal website quality or that screenshots and browser telemetry alone establish whether a website is good.

The core observation is seller-side production proof. Separate later evidence is required for unrelated external purchasing, autonomous seller selection and independent Assurance.

Why it mattered to later Atinamos work

Render Check made the first problem tangible: a useful service could be expressed in a form another machine could buy. Buyer experiments then exposed the next problem — a technically functional service may still be overlooked if its machine-facing contract is unclear.

That work eventually led beyond discovery and payment into the harder question that now defines the Labs research programme: what independent evidence should an autonomous buyer use before deciding to spend?

Public source

The original public case study retains the Render Check workflow, browser evidence examples and seller-side production proof.

Open the Atinamos agent-commerce case study ↗