Your docs are killing AI agents.
We show you where.
Your customers point Claude Code, Cursor and Codex at your docs. Those agents get lost, invent endpoints and give up — and you never hear about it. We send 6 in and film every failure.
A checker tells you llms.txt is missing.
We show you the agent that died on page 3.
Meet your first Casualty.
A Dummy with one job — make a first successful API call — works through a real docs site until it can't. The screen stamps the exact URL where it gave up.
Sample run shown above. Real Dummy, real docs page — yours will point somewhere else.
You wrote your docs for humans, and they work for humans. Then half your readers turned into agents nobody tested for. An agent that gets lost does not file an issue. It tells your customer your API does not do that.
One purchase. The whole gauntlet.
No dashboard to configure. No score to interpret. You pay, the agents run, you watch.
Fund the run
$7, paid before anything moves. Hosted Stripe checkout, NanoCorp as merchant of record. Every run is funded — nothing runs on the house.
Release the Dummies
6 real agents, 6 jobs your customers actually ask theirs to do: make a first API call, answer a support question, sign up, find your spec.
Watch the Casualty
The first Dummy dies inside 30 seconds. A Cause of Death names the page, the status code, or the parameter it hallucinated.
Take the prompt
You leave with the fix, not a grade: the agent-facing files plus one long prompt for Claude Code, Cursor or Codex. Paste, ship, run it again.
Questions before you run one
The opposite. We do not touch your site or change anything on it. We send agents in to find where they fail, so the ones your customers send can get through.
An agent is reading your docs right now.
It will not tell you how that went.
$7 buys 6 Dummies, the page that breaks them, and the prompt that fixes it.
We read your site first and show you the squad before you pay.