Agent-Ready score

Can a buyer’s agent buy your product?

Ten checks across the five tests an agent runs on your product, in order. Two minutes. The result names the test where agents stop, which is your first project.

Score your product → Read the essay
The score

Ten checks, five tests.

Answer for your product as it is today, not the roadmap. “Partly” means it works with a human from your side in the loop.

1 Discoverable Do agents name you?

Ask ChatGPT, Claude and Gemini to solve your core job. Do they name your product?

Are your docs public, with an llms.txt or an MCP server an agent can find?

2 Evaluable Can an agent compare you without a sales call?

Is your pricing, with its limits, published somewhere an agent can read it?

Can an agent get sandbox or trial keys without talking to anyone?

3 Transactable Can an agent sign up and pay?

Can an agent create an account and get API keys programmatically?

Can it start paying (card, usage billing or agent checkout) with no one from your side involved?

4 Operable Can an agent do everything a user can?

Can everything a user does in your UI also be done through your API or MCP server?

Do your errors tell the caller how to fix the problem, not only that it failed?

5 Trustworthy Would a CISO let an agent in?

Can a customer give an agent scoped permissions and a spend cap?

Do agent actions show in audit logs, with an approval step for risky ones?

0 of 10 answered

Your answers are scored in your browser and are not stored. Leaving an email for the written version is optional.

Why order matters

An agent stops at the first test you fail.

A person who hits a sales-call wall might book the call. An agent moves on to the next vendor and never tells you it was there. That is why this score is not an average: a great API does not help if the agent could not sign up to reach it. The five tests and what good looks like for each are in Agent-Led Growth.

Talk through your result →

Questions this raises

What does Agent-Ready mean?

A product is Agent-Ready when an agent acting for a buyer can get from “I need this job done” to “done” without a human from your company. It comes down to five tests, run in order: discoverable, evaluable, transactable, operable and trustworthy. The full model is in the essay Agent-Led Growth.

Why does the score stop at the first failed test?

Because that is what a buyer’s agent does. An agent that cannot sign up never reaches your API, however good the API is, and it does not ask you for help. So the result names the first test you fail, and that test is your first project.

What comes after Agent-Ready?

Agent-Run: your own agents carrying customers from first touch to expansion, with people on judgment, trust and signatures. Start with activation, not outbound. The 90-day plan in the essay sets the order and the gate between the two.