The agent successfully gathered enough information to explain Kabeli's core purpose, target audience, and differentiation philosophy, though it hit hard stops on pricing and specific product details. The site is moderately agent-ready: it publishes structured context via llms.txt and llms-full.txt, maintains a philosophy-driven blog, and provides text versions of key pages, but deliberately obscures pricing and concrete deliverables, requiring the agent to assemble a portrait from fragments and infer positioning from blog posts.

• ›Steps [7], [8], and [18] (the .txt artifacts) were the most content-bearing — they contained the company description, service categories, target audiences, and contact info in structured, machine-readable form. These were discovered via prior knowledge / direct guessing of common artifact paths, not surfaced by navigation.
• ›Steps [15] and [16] (blog posts 'Built for Institutions, Not Demos' and 'AI That Earns Its Place') were essential to understanding differentiation, but required the agent to extract philosophy and inference from narrative content rather than finding a dedicated 'why we're different' page. The RSS feed [13] signaled their existence but not their content.
• ›Pricing remains deliberately inaccessible: /pricing returns 404, and no public rate card exists. The agent inferred a consultative, case-by-case model from the 'Book a Demo' CTA, but this was educated guessing rather than explicit disclosure. The site prioritizes principle-driven positioning over transactional clarity.
• ›The site does not expose concrete deliverables, case studies, team info, or project examples in a machine-readable form. The agent had to rely on blog philosophy to infer what 'institutional software' and 'applied AI' mean in practice, leaving ambiguity about whether they are a services firm, product vendor, or hybrid.
• ›Text artifacts (.txt files) suggest the site was designed with AI agents in mind, but they are supplementary to HTML pages rather than the primary source of truth. The HTML pages are React/Next.js bundles that render minimal semantic content in the raw responses, making agent extraction harder without full rendering.