The agent assembled a coherent explanation of TerraConstructs by combining the GitHub README (step [11]) with inferred positioning from web search snippets and prior knowledge, but the main website itself was largely inaccessible — it's a Next.js SPA that returned only boilerplate HTML without rendered content. The agent successfully answered what TerraConstructs does, who it's for, and how it differs from alternatives, but had to work around a non-discoverable pricing model and acknowledged significant gaps in public documentation.

• ›Step [11] (GitHub README) was the only content-bearing source that provided concrete technical details: the library's core abstractions (asset pipeline, IAM grants, Step Functions), its foundation on CDK Terrain, and comparison to raw HCL and CDKTF. This was the backbone of the final answer.
• ›Step [9] (GitHub repo page) was cited as a source but returned only HTML boilerplate; the actual README had to be fetched separately (step [11]) to extract usable information.
• ›The main website (terraconstructs.dev, steps [1] and [3]) returned truncated Next.js HTML with no rendered page content visible to the agent. Targeted paths for pricing and docs (steps [5–6]) returned 404s, forcing reliance on GitHub and web search.
• ›Web search results (steps [7], [13–14], [18–19], [21], [23]) were routers only — they confirmed the domain and linked to GitHub, but the search snippets themselves contained no differentiating details. The agent had to use prior knowledge to construct the pricing model ('completely free, open-source') and positioning (Pulumi is proprietary, CDKTF is deprecated).
• ›Pricing information was not published on the site or GitHub; the agent inferred it from the open-source license and lack of any commercial messaging, then confirmed it against prior knowledge that CDKTF was deprecated in December 2025.