How ready is this product for AI agent commerce?
When AI agents start buying things, which businesses will they be able to work with? That turns out to be a surprisingly hard question to answer by just looking at a website.
Some things are obvious. If a product has a public API with documentation, that's a good sign. If the only way to buy is "contact sales," that's a bad sign. But most companies fall somewhere in the messy middle — they have an API but no machine-readable pricing, or they offer self-serve signup but gate it behind a CAPTCHA.
We wanted a way to be more systematic about it. Not because a score can capture everything, but because the exercise of defining what "agent-ready" means forces you to be specific about what actually matters.
We score four categories. Each is worth 10 points, for a total of 40.
Can an agent find this product and understand what it does?
If an agent can't figure out what you sell and what it costs without parsing marketing copy, you're invisible to the machine economy.
Can an agent buy this without a human?
Every "contact us for pricing" page is a door that's closed to agents.
Can an agent actually use this?
The difference between a product an agent can use and one it can't often comes down to whether there's a test environment.
Can an agent verify this is reliable?
When agents are spending money autonomously, they need programmatic ways to assess risk — not a logo wall of enterprise customers.
No. A 40-point automated crawl is a blunt instrument — more conversation starter than definitive assessment. But even a rough score surfaces patterns that are hard to see otherwise.
We're putting this out there not because we think the scoring is perfect, but because we think the conversation matters. What should agent-readiness mean? Are we weighting the right things?
Tell us what we got wrong — on X or in the Substack comments.