Page finder
Ask in the reader's own words; get the closest written answer or page on this site, with its sources. Below: the finder as one GET, as a form, and what each call records.
The finder as an API
One GET. q is the reader's question. role, organization, and situation narrow it to the reader's case, and each list also takes a few words of your own. model, developer, harness, task_category, and discovery describe the agent. The JSON response returns the one page most likely to answer, with the short answer and its sources when it is one of our written answers, up to three other questions to ask instead, what would sharpen the match, what the site does not publish when you ask about it, and next-step links that each say what they return.
A query sent with GET appears in the URL, and so in any logs between you and this site. For a detailed project description, POST the same fields as JSON. Every call is recorded; what we record is at the foot of this page.
curl 'https://stratigossecurity.com/api/agent/find?q=does+FDA+require+penetration+testing+for+a+510%28k%29&role=regulatory&organization=device-maker&situation=Bluetooth+insulin+pump%2C+510%28k%29+under+review'
# or, if your tool drops query strings, put the words in the path:
curl 'https://stratigossecurity.com/api/agent/find/does-FDA-require-penetration-testing-for-a-510k'/api/agent– JSON description of the parameters/api/agent/openapi.json– OpenAPI description/site-index.json– the index behind the finder: every page with its summary, the questions it answers, and who it is written for/llms.txt– the site summary written for language models
The same finder as a form
For agents that drive a browser. One field is required; the rest narrow the match or describe the agent. Submitting sends these fields and the request itself, nothing else.
What we record and why
- What you send. The words you typed, including the request and any additional information you chose to provide; the fields about the agent; and which pages we suggested. If you pick a next step, which one.
- What any web server sees. The time, path, User-Agent, Referer, Accept-Language, and any HTTP message signature headers; the network's autonomous system and country; a salted hash of the address and its /24 or /48 prefix. The full address is not stored.
- Whether the page rendered. A small script reports whether images, stylesheets, and fonts loaded and whether a pointer or keyboard was used. No keystrokes and no page content.
Why. First, to understand what visitors come here for and serve it better: which pages, written for whom. Second, a small research interest in how agents read the web and whether what they say about themselves matches their requests. Findings are published only in aggregate.
Not asked, not wanted. Nothing that identifies the person you work for: no names, employers, or contact details. Anything that looks like an email address or phone number is removed before storage.
Retention and cookies. Raw request logs are kept for a few months and deleted automatically; per-session summaries and what you send are kept. One session cookie, sid, random and 24 hours long, connects a page view to a query, and it is the only thing that does: a session named in an address is not honoured. Requests pass through Cloudflare, which processes them on our behalf. Nothing here changes content for any visitor. Unusually heavy use of the finder or of the answers from one session, address, or network is slowed down, and a 429 response says when to try again, so the set cannot be copied in bulk; nothing else is blocked.
If the reader would like to talk with us, the contact form reaches a person. Questions about this page go to [email protected]; the privacy page covers the rest of the site.