ora
metal.ai
success
Claude Code
intent

What does metal.ai do and who is it for? Explain it back to me.

20steps
102.7sduration
$0.2729cost
284,117tokens
23 steps3 reasoning steps8 searches
home
docs
docs
/sites/6wwgkwjzsneop36…
home
docs
/solutions
search
docs
search
/company
search
/product
search
docs
search
search
search
docs
search
58%
on-site discovery
67%
reliability
50%
link following
path origin
  • previous resource50%
  • web search42%
  • prior knowledge8%
insight

The agent successfully assembled a comprehensive explanation of Metal.ai's core business, target market, features, and competitive positioning by combining the homepage, product pages, and blog articles. However, pricing information could not be retrieved from the site itself—only a custom-quote requirement was stated—forcing the agent to explicitly flag this as missing. The site is moderately navigable for discovery (blog and product pages exist and loaded) but has limited machine-readable structure; most content was extracted from JavaScript-rendered pages and fragments rather than published as structured data.

  • ›Steps [1], [16], [17], and [20] provided the core content: the homepage established the core value prop (Context Graph for deals/documents/decisions), the /product page articulated the differentiator vs. RAG, and the /blog aggregated case studies and technical explanations. These pages loaded successfully and contained substantive information.
  • ›Steps [12] and [14] (blog posts 'How Context Graphs Are Replacing RAG' and 'Deal Intelligence') were the primary sources for explaining the competitive differentiation and how the product works—the agent had to navigate to blog content via search because the homepage did not directly surface detailed feature explanations.
  • ›Pricing is completely absent from the site. The agent attempted /pricing (returned 404), and Metal only states 'contact for quote' on public pages. The agent had to explicitly call out this gap and could not fulfill that part of the task from the site alone, despite it being part of the core request.
  • ›The site is JavaScript-heavy (Framer-rendered) and does not expose structured metadata or API documentation publicly. The agent had to guess at several URLs (/pricing, /about, /docs, /solutions all returned 404 or redirects to the homepage), indicating weak URL structure and reliance on runtime JavaScript rendering.
  • ›The agent successfully inferred company background (founders, Y Combinator, funding) and customer names (Clearlake, Berkshire Partners, Blue Wolf Capital) from fetched pages and citations, but had to cross-reference Y Combinator and VentureBeat articles to fill gaps the main site did not expose.

Want to run your own?

Join the waitlist for early access to point your own agents at any domain, with the intents you choose.

or talk to us about agent readiness →