The agent successfully compiled a comprehensive explanation of Arcade.dev by assembling information from the site's own content (homepage, llms.txt, llms-full.txt, product page) and third-party analyses. The site exposed core positioning and capabilities well through structured text files, but pricing details were obstructed by JavaScript rendering, forcing the agent to rely on external sources and acknowledge gaps rather than deliver fully transparent pricing information.

• ›Steps [5] and [6] (llms.txt and llms-full.txt) were the most content-bearing: they provided clear, machine-readable positioning ('enterprise-ready actions runtime'), three-pillar capability breakdown (authorization, tools, governance), and use cases. These artifact probes worked as intended.
• ›Step [0] (homepage) and step [10] (product page) returned HTML with meta descriptions but the agent could not extract full feature/pricing details from the rendered page structure—indicating the site relies on client-side rendering for detailed feature matrices.
• ›Pricing information (steps [1] and [8]) failed to yield actionable detail despite being fetched twice. The agent was forced to search externally (steps [7], [14], [15]) and cite third-party sources (Kingy AI, Scalekit, The AI Agent Index) to answer the pricing question. The agent explicitly flagged this opacity in its 'What Was Confusing' section.
• ›The agent successfully differentiated Arcade from competitors (Composio, Truto, Portkey, Unified MCP) by pulling from a dedicated external comparison article [13], demonstrating the site itself does not publish a structured alternatives/comparison resource.
• ›The site is agent-ready for discovery and identity (clear tagline, positioning, capability pillars, 8,000+ tools, integrations list all surface), but poor for payments layer (pricing page is JavaScript-heavy and non-machine-readable).