SerpAPI

**Judgement:** Recommended as a strong, general-purpose web and product/local-search tool, especially when an agent needs fresh, structured data from multiple major search engines. **What it does well** - Wraps SerpAPI into a simple script (`serp.py`) that supports many engines (Google, Maps, Shopping, Images, News, Amazon, Yelp, OpenTable, Walmart, eBay, Tripadvisor, etc.). - Returns structured JSON with rich fields (prices, ratings, reviews, GPS coordinates, knowledge-graph data), ideal for downstream reasoning and aggregation. - Supports key options (location, number of results, pagination, language, country, output format) that are important for localized and targeted queries. **Key risks & limitations** - **External dependency:** All functionality depends on SerpAPI and the underlying search engines; outages, API changes, or engine-side anti-bot measures can degrade or break behavior. - **API key and quotas:** Requires `SERPAPI_API_KEY` and is subject to SerpAPI rate limits (e.g., 100/month free tier). Agents may hit limits unexpectedly under heavy use. - **Content quality & safety:** Results are unfiltered web content (news, general web, images, product listings, reviews). Agents must not treat outputs as verified facts and should apply their own safety, bias, and misinformation checks. - **Jurisdictional/ToS constraints:** Different search engines and locales may have legal or policy restrictions on scraping/aggregation; usage should comply with SerpAPI and source-engine terms. **Best-fit scenarios** - Tasks that need **up-to-date web knowledge** beyond the model’s training cutoff (e.g., current events, recent products, live inventory/price checks). - **Local discovery and recommendation** flows: “find the best X near Y,” with ratings, hours, and coordinates for mapping or route planning. - **Shopping and price-comparison** workflows where an agent compares offerings across Amazon, Google Shopping, Walmart, eBay, etc. - **News and monitoring**: tracking headlines and coverage patterns around a topic or entity in near real time. Use this skill when an agent needs real-time, structured search data across multiple platforms via a single, well-documented interface, while layering its own verification and safety logic on top of the raw results.