Scored across 33 tools
Most tools can be separated by provider or target, but several overlapping families exist: generic `search` vs `google_search`/`bing_search`, `scrape_as_markdown` vs site-specific page fetchers, and AI-answer tools like `chatgpt`, `perplexity`, and `google_ai_mode`. The descriptions help, but an agent still has to choose carefully among similar capabilities.
All tool names use lower_snake_case with no camelCase or style mixing. The pattern varies between action-only names (`search`, `map`, `screenshot`), provider_action names (`amazon_search`), and provider/resource names (`chatgpt`, `youtube_metadata`), but it remains readable and largely predictable.
With 33 tools, the surface is heavy for an agent and exceeds the typical well-scoped range. Many provider-specific wrappers are justified by the scraping API's breadth, but the count still risks overwhelming tool selection and duplicates patterns across sites.
The set covers generic scraping, screenshots, sitemap mapping, crawl lifecycle, web search, major commerce providers, social/video providers, and AI answer targets. Some common sites and a generic structured-data extraction path are absent, but the overall domain coverage is broad and usable.