li.fi
success
Claude Code
intent
What does li.fi do and who is it for? Explain it back to me.
8steps
45.7sduration
$0.0494cost
27,976tokens
11 steps3 reasoning steps
home
docs
docs
docs
home
/plans
home
/use-cases
80%
answer from your site
50%
answer efficiency
86%
followed site links
answer sources
- from this site80%
- from agent knowledge20%
url discovery
- given in the task13%
- followed a link75%
- guessed the URL13%
insight
The agent successfully assembled a comprehensive explanation of li.fi's core business, pricing, and differentiation from the homepage and pricing page alone. The site's public-facing content was navigable and information-dense, but critical URLs (docs, about, use-cases) returned 404s, forcing the agent to rely on what the homepage and pricing page explicitly stated; notably, the agent correctly identified gaps in the published information (enterprise pricing details, fee structure specifics, technical architecture) rather than fabricating them.
- ›Steps [1], [6], and [9] all fetched the homepage (/) but returned different response summaries, suggesting the agent performed multiple passes or the LLM extracted different subsets of content each time—all key claims in the final response trace back to these homepage fetches.
- ›Step [7] (pricing page) was the only successfully fetched non-homepage URL and was the sole source for the Standard/Enterprise tier breakdown; this page was critical to answering the pricing question and was correctly identified and followed by the agent.
- ›Multiple 404s ([2] docs, [3] pricing attempted before success, [4] use-cases, [5] about) indicate the site's navigation structure does not expose these as direct links, forcing the agent to guess URLs—the agent recovered by re-fetching the homepage and following the pricing link when it appeared in the response.
- ›The final response's 'Things I Couldn't Find' section ([1] enterprise pricing specifics, [2] fee model details, [3] competitor comparisons, [4] rate limit specs, [5] technical architecture) are all claims the agent correctly attributed to site gaps rather than model memory, demonstrating honest gap-reporting.
- ›The site content was sufficient to answer 'what they do,' 'who it's for,' and 'how it's priced' directly, but the agent had to infer the differentiation structure (meta-aggregator vs. direct competitor positioning) from homepage language about 'single integration' and 'connecting to' other protocols—not a separate 'Alternatives' or 'Comparison' section.
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 →