Agent successfully gathered comprehensive information about LokI's value proposition, target market, and pricing tiers by fetching the homepage, pilot page, and hardware page. The site is moderately navigable for this task, but lacks a dedicated pricing page, docs, or about section; key details like software licensing terms, underlying model identity, and post-pilot progression are vague or absent.

• ›Pricing is fragmented across three pages (homepage, /pilot/, /hardware/) and split into three components (hardware, installation, custom software quote), with no single comprehensive pricing resource. Software cost remains entirely opaque, forcing users to request a quote.
• ›Homepage stats dashboard shows zero values with no explanation (0 test cases, agents, passing scores, files sent to cloud), undermining credibility claims like '100% logged' and '0 files to cloud' — these appear to be unsupported or placeholder content.
• ›The underlying AI model is never named; references to 'groot model' for nightly training and the inference framework (MLX, llama.cpp, vLLM on hardware pages) are mentioned without disclosure of which model powers the agents.
• ›Agent correctly identified that the site surfaces core facts (what it does, pilot phases, hardware tiers) in readable form across navigable pages, but the site requires explicit demo-request contact to clarify software licensing, total cost of ownership, and post-pilot commitments.
• ›Server tier pricing is marked 'indicative' with ranges (e.g., €9,000–12,000) rather than fixed prices, and no recurring subscription or renewal fee structure is disclosed, creating cost-of-ownership ambiguity the agent noted but could not resolve from available pages.