@pipeworx/bonhams

Connect: https://gateway.pipeworx.io/bonhams/mcp · Install: one-click buttons

Tools: 2

Realized auction prices from Bonhams — search past (“sold”) lots by artist or maker and get the hammer price, the hammer price including buyer’s premium (the actual realized/sold price), estimate range, sale date and department; pull full lot detail (description, condition, images) for one lot.

Tools

• results_search(query, category?, page?, limit?) — search Bonhams’ sold-lot archive by artist/maker/keyword. category is a free-text filter matched against the department name (e.g. “jewellery”, “prints”, “books”).
• lot_details(auction_id, lot_no, slug?) — full detail for one lot, keyed by the auction_id/lot_no a results_search result carries.

Auth

Keyless. Public pages, no login required.

Data sources

• https://www.bonhams.com/search/?chronology=past&query=<q>&page=<n> — the site’s own past-lots search page. It’s a Next.js app that server-renders the FULL result set into a __NEXT_DATA__ JSON blob (pageProps.lotData.auctionLots[]) — a plain fetch() gets real data, no separate XHR/API call needed. Each entry carries price.hammerPrice (before premium) and price.hammerPremium (the realized price a buyer actually paid, matching the “Sold for $X inc. premium” text shown on the page) plus estimateLow/estimateHigh, department.name, hammerTime.datetime, auctionId, and lotNo (which is {full, letter, number} — full is sometimes absent, fall back to number).
• https://www.bonhams.com/auction/<auctionId>/lot/<lotNo>/<slug>/ — the lot detail page, same SSR pattern, one "lot":{...} object with richer fields: sDesc/sCatalogDesc (HTML description), dHammerPrice, dHammerPremium, dEstimateLow/dEstimateHigh, images[]. The slug in the URL is not validated — any string works as long as auctionId/lotNo match, so a caller doesn’t need Bonhams’ exact descriptive slug.
• robots.txt (checked 2026-09-05) disallows only /ldc/, /vms-assets/, *aggregate$, *head_image* — none of which this pack touches. No crawl-delay declared for User-agent: *.

Survey notes for the other two houses (Sotheby’s, Phillips)

Bonhams was shipped first because its realized prices are plain SSR HTML with no gate — the easiest shape found. Quick probe notes on the other two, recorded for whoever builds their packs next:

• Sotheby’s — robots.txt sets Crawl-delay: 15 for User-agent: * and disallows /bsp-api/* (their actual API path) plus all PDF exports. Results pages are likely a genuine client-side API call behind /bsp-api/, which robots explicitly blocks — worth checking whether a different, non-disallowed endpoint serves the same data before assuming it’s blocked outright.
• Phillips — robots.txt disallows /Search, /search, /SEARCH, /bin/, /phillips/otis, /*/filter/. The obvious keyword-search path is blocked case-insensitively; a lot-detail page reached some other way (e.g. from a sitemap or a department listing) may still be fair game — not probed in this session.
• Christie’s — out of scope per the build plan (closed, internal API).

Bonhams proved the per-house-pack shape works, so follow-on fleet tasks were filed for the other two: #1280 (Sotheby’s) and #1281 (Phillips), each carrying the probe notes above.

Tools

• results_search — Search Bonhams’ past-lot realized-price archive by artist/maker (or any keyword). Returns sold lots with hammer price, hammer price including buyer’s premium (the actual realized price), estimate rang
• lot_details — Full detail for one Bonhams lot: description, condition, images, realized price (hammer + buyer’s premium), estimate range, sale date. Needs the auction_id and lot_no from results_search (slug is opti