Some cookies are necessary for SecondDoor to work. Others are optional, and tell us how the site and the product are used. Accept all, or decline all optional ones. Without a selection, nothing optional is set. More in our cookie policy.
Basic scanSoftware business · Marketing site/Scanned 8 Sept 2026, 14:49 UTC/1 browser pass
A product with published pricing and a one-field signup, but no API documentation or self-serve credentials for agents to integrate.
Older scan
This report was produced by an earlier version of the scanner. Some findings may not match how the site is measured now.
Present. Agents lose ground at specific gates, not everywhere.
Ordered by what the evidence says is costing most. Written for whoever owns the site, not whoever builds it.
1 piece of evidence and the recommended fix are recorded for this issue, with /compare/profound-vs-competitor written from your own pages.
This is a basic scan, so it raises the three issues the deterministic evidence supports. A deep scan sends two agents over the site and raises up to ten, each with cited evidence and the screens to prove it.
A deep scan walks the site with two agents, records cited evidence and screenshots, and keeps a mission timeline. It comes with the paid plans. Making an account costs nothing and gives you one basic scan a month.
The score and the audit trail stay public. The evidence and the files open with an account.
The audit trail behind every number above. Not measured is not zero.
Machine discoverability
No AI crawler is blocked by robots.txt, a real browser is served the page normally, and the sitemap lists 411 URLs. However, the homepage carries no schema.org markup and /llms.txt is not published, so machines cannot infer what the.
4 observations recorded.
Programmatic onboarding
The documentation site at docs.tryprofound.com contains 2634 characters of text but shows no API credentials, endpoints, code samples, or authentication documentation.
3 observations recorded.
Pricing legibility
8 prices are readable in the homepage HTML without JavaScript, across three named tiers: Starter, Growth, and Enterprise. All prices are published; the Enterprise tier is not gated behind a demo request or sales contact form.
1 observation recorded.
Agent-aware instrumentation
An MCP manifest is published at https://docs.tryprofound.com/.well-known/mcp.json, enabling Claude and other MCP-aware agents to discover the service. No OpenAPI spec, ai-plugin.json, or OAuth authorization metadata is published.
4 observations recorded.
Machine-fetchable trust
security.txt is published at /.well-known/security.txt with a contact route for machines. The homepage responds to automated requests with HTTP 200 and serves content normally to a browser carrying an AI user agent.
2 observations recorded.
Commercial rails
Prices are published in raw HTML on the pricing page with no JavaScript required. A self-serve signup form is reachable from the homepage via the Get Started button.
2 observations recorded.
Steps to first value
The Get Started button on the homepage leads to a signup form in 2 steps. The form asks for one field (email) and carries no payment requirement, so a visitor can create an account and access the product without friction.
1 observation recorded.
Required fields
The signup form contains 1 field total, and 1 of 1 is required (email). No optional fields, phone numbers, company names, or verification codes are requested before account creation.
1 observation recorded.
Verification walls
The signup page does not display email verification, SMS verification, or CAPTCHA requirements in the form itself. The walk could not observe what happens after submission, so verification may occur after the form is posted.
1 observation recorded.
Error recovery
All 1 input field on the signup form carries a validation rule (email format). No inline error containers or live regions are present to guide users through correction, but the validation attribute itself enables browser-native error.
1 observation recorded.
It got through 8 of 15.
The agent got through 8 of 15 requests, but could not read the site's own guide for language models. It came up empty on 6 other checks too.