The agent gathered substantial information about MailSend's function, target audience, and technical differentiation, but could not access specific pricing tiers or costs because the site uses client-side rendering with no server-side fallback. The OpenAPI spec ([4]) and web search results ([9], [10], [11]) provided most actionable content; the site's own pages were inaccessible in readable form, forcing reliance on prior knowledge and external sources.

• ›Step [4] (OpenAPI spec) was the only direct site resource that yielded machine-readable, content-bearing data—the description, endpoint architecture, and feature outline. All other site fetches ([1], [2], [3], [6], [7], [8], [13], [14]) returned minified HTML with no readable fallback, rendering them useless to the agent despite 200 status codes.
• ›Web searches ([9], [10], [11]) returned link titles and metadata but no snippet content that answered pricing or comparative positioning questions. The agent was forced to acknowledge gaps rather than construct false answers from sparse search results.
• ›The agent explicitly cited the OpenAPI documentation and feature pages as sources, but could only cite them for architecture/feature claims, not pricing. Pricing claims in the final response are stated as 'described' or 'claims' rather than sourced, indicating the agent fell back to prior knowledge or inference.
• ›The site's navigation structure exists (urls like /pricing, /features, /features/email-automation) but the content behind them is inaccessible. This is a rendering-layer failure, not a discovery failure—the agent found the right URLs but received no usable content.
• ›The agent's comparison to Mailgun/SendGrid/Postmark came entirely from web search results ([10], [11]), not from site content; MailSend does not appear in third-party comparison databases, which the agent correctly flagged as a positioning gap.