Scraping Amazon without controlling location produces numbers that are true somewhere and wrong for the customer you care about — availability, delivery estimates and sometimes price all move with the postcode. Exposing that control is the difference between usable data and plausible data. The offer-listing tool is the other one worth calling out, since the buy-box price is one seller's and the full offer list is what actually tells you the market. proxy_mode_config generating configuration without executing is a small courtesy that avoids paying for a request just to learn the parameter format.
A Scrape.do client covering three modes: direct scraping with structured extractors for Google and Amazon, an async job API for volume, and a proxy-mode configuration generator.
- scrape_url returns Markdown or raw output, so a page can arrive in a form a model reads rather than as HTML to parse.
- google_search returns structured SERP data with localisation support rather than a scraped results page.
- Amazon covered at four levels — product detail, all seller offers with pricing and fulfilment, search and category pages, and raw HTML with ZIP-code geo-targeting for the cases the structured extractors do not cover.
- The async API is a full job system: create, list with pagination, read job and task detail, cancel, and read account concurrency limits.
- proxy_mode_config generates the configuration and parameter strings without executing anything, so the setup can be produced without spending a request.
SCRAPE_DO_TOKEN. Requests are metered, and the async account tool reports the concurrency you are entitled to.
One command plus a key — claude mcp add-json scrape-do --scope user '{ "type": "stdio", "command": "npx", "args": ["-y", "scrape-do-mcp"], "env": { "SCRAPE_DO_TOKEN": "YOUR_TOKEN_HERE" } }', then supply credentials