For teams building APIs and developer tools
I test how agents choose and integrate your product. You get the steps that failed, the evidence, and a specific change to try.
Krystian Gwizdała · I run the tests and discuss the results with you.
Recommendation studies include Claude Code, Codex, Cursor Auto and Antigravity. Tool coverage and dates are shown in each report.
Choose what you want to check
Can an agent read how to get started?
16 checks of your public docs, signup pages and API access instructions. Each result links to what the scanner found.
HTTP checks only. No agent tries your signup or integration.
Your scan gets a permanent link you can forward. It is never added to the published corpus, and we do not post it anywhere.
No account or card required.
Does your product make the shortlist?
One buying question, at least ten answers across two tools. See which providers get named and read every answer.
One report, paid once. I check that the question fits your product before you pay.
Read a sample reportCan an agent finish the job?
Two tasks, two tools. I test the integration, make one small docs or example fix, then run the tasks again.
$1,500 each for the first 2 pilots. Up to 12 hours of my work. A retest may show no improvement.
See the pilot scopeThe score covers public HTTP signals across five stages. Unmeasured and inapplicable checks are excluded from your denominator.
Across 181 published domains, an average of 14 of 18 possible points could be measured. Only 1 could be measured in full.
80 signup pages returned no form without JavaScript. That alone does not establish a blocked signup: a browser-capable agent may still use them.
Read all checks and scoring rulesFree monitoring beta · available on its own
Weekly public-site checks. Confirmed changes arrive with the old result, the new one and a link to the evidence.
No audit needed. I will ask before charging. Unsubscribe from any email.
Agent mention checks are a separate manual experiment, with no fixed schedule or guaranteed run count. They cover one buying question in each of 26 categories, without signup or integration.
Published research
0 of 12
Across the three categories that needed credentials, all twelve runs required human help to obtain them.
See the integration findings ↗5/8 → 0/12
One provider was chosen in five of eight empty-project runs, and none of twelve runs with an existing app.
Compare the two conditions ↗These are independent studies, not paid client results. They do not measure lost sales or prove that a fix will change an agent's choice.
181 domains in 26 categories large enough to rank · every result is a live link
Ordered by the share of measurable points earned. Unmeasured and inapplicable checks do not count against a vendor.
An asterisk (*) means too little was measurable for a fair comparison. These rows appear last, regardless of their percentage.
8 more in this category, all of them in the data.
5 more in this category, all of them in the data.
4 more in this category, all of them in the data.
3 more in this category, all of them in the data.
2 more in this category, all of them in the data.
2 more in this category, all of them in the data.
2 more in this category, all of them in the data.
2 more in this category, all of them in the data.
2 more in this category, all of them in the data.
2 more in this category, all of them in the data.
2 more in this category, all of them in the data.
2 more in this category, all of them in the data.
2 more in this category, all of them in the data.
1 more in this category, all of them in the data.
1 more in this category, all of them in the data.
1 more in this category, all of them in the data.
1 more in this category, all of them in the data.
1 more in this category, all of them in the data.
1 more in this category, all of them in the data.
1 more in this category, all of them in the data.
1 more in this category, all of them in the data.
1 more in this category, all of them in the data.
1 more in this category, all of them in the data.
1 more in this category, all of them in the data.
1 more in this category, all of them in the data.
1 more in this category, all of them in the data.