rakama.app
partial
Claude Code
intent
What does rakama.app do and who is it for? Explain it back to me.
24steps
104.6sduration
$1.2261cost
509,224tokens
44 steps20 reasoning steps7 searches
home
home
docs
/features
docs
/faq
search
search
search
search
/help
docs
docs
docs
docs
search
search
/privacy
/contact
sitemap
docs
search
/terms
docs
100%
on-site discovery
65%
reliability
47%
link following
path origin
- previous resource47%
- prior knowledge53%
insight
The agent gathered a partial understanding of Rakama by assembling information from the homepage, blog posts, and sitemap, but could not complete the task due to inaccessible pricing information. The site's HTML responses were truncated and heavily client-side rendered, forcing the agent to rely on prior knowledge and blog post titles to infer features; the critical pricing page was either not published or behind interactive elements the agent could not fully retrieve.
- ›Steps [1], [19], [29], [37], [42] returned truncated HTML with 'truncated, 4400 more chars' markers, indicating the fetches captured only metadata and structure, not full content. The agent had to piece together understanding from page titles (e.g., 'how-to-split-a-bill-in-rakama', 'what-is-a-financial-health-score') rather than actual body text.
- ›Step [35] (sitemap.xml) revealed the site structure and confirmed the existence of a pricing page, but steps [5] and related pricing lookups returned 404s or inaccessible pages. The homepage clearly references pricing in navigation (title mentions 'Track. Budget. Achieve.'), yet the agent could not retrieve its full content despite multiple attempts.
- ›The agent relied heavily on prior knowledge (53% sourced from prior_knowledge per run metadata) to construct claims about features like 'automatic categorization and AI', 'bill splitting logic', and 'no advertising/analytics'. These appear in the final response but were not confirmed in the trajectory's fetched content—they likely came from the agent's training data about rakama.app rather than live retrieval.
- ›Blog posts were the most valuable discovery surface ([22], [29], [37], [42]): the sitemap led the agent to specific blog URLs that hint at features. However, the actual blog body content remained truncated, so the agent could only use the URL slugs and inferred titles as signals.
- ›The task explicitly asked for pricing ('how it's priced'), and the agent marked this as "⚠️ UNCLEAR" and "the most confusing aspect", acknowledging task incompleteness. No pricing tiers, costs, or free/premium split was retrieved from the site.
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 →