Build AI agents on luxury watch data — Vibecoder kit

One-line install

{"mcpServers":{"mostexpensivewatches":{"url":"https://mostexpensivewatches.net/mcp"}}}

Streamable HTTP · JSON-RPC 2.0 · protocolVersion 2024-11-05 · free, no auth.

Workflow 1 — Watch-buyer copilot

Your user says "I have $80k to spend on a Patek Philippe perpetual calendar, what's out there?" The agent calls search_inventory with brand: "Patek Philippe", complication: "Perpetual Calendar", max_price_usd: 80000, gets up to 8 live listings with merchant URLs, then calls get_market_estimate on the top match to surface "is this price reasonable vs auction comps?" Hand the answer to the user with citations.

Example prompt to paste into Claude / Cursor with MEW installed

You are a watch-buying copilot. The user has $80k to spend on a Patek Philippe perpetual calendar. Use the mostexpensivewatches tools:
1. search_inventory for in-stock Patek Philippe perpetual calendars under $80,000.
2. For the top 2 hits, call get_market_estimate to compare asking price to auction comps.
3. Recommend the one with best value-to-comp ratio. Always include the listing URL and the auction record URL the agent cited.

Workflow 2 — Auction tracker

Your user wants to be alerted when a specific reference shows up at upcoming sales. The agent calls list_upcoming_auctions (optionally filtered by house) and matches the highlights against the user's watchlist. For each match it calls search_auction_comps to surface the historical hammer range and inform a bidding ceiling.

Example prompt

Background watchlist: ["Patek 5711/1A", "Patek 5970", "Lange Datograph"]
Use mostexpensivewatches.list_upcoming_auctions, then for each upcoming sale, check if its highlights mention any watchlist item. For each match, also call search_auction_comps for that reference and report:
  - sale date, house, city
  - highlight description
  - historical hammer median + record
  - lot URL
Reply with a short table.

Workflow 3 — Market-research agent

Your user is researching a brand or reference. The agent calls lookup_reference to fetch the AI-written factual description + listing stats + auction record. Then get_brand_info for context. Then search_answers for any published explainers. Hand a synthesised report with citations.

Example prompt

Research the A. Lange & Söhne Datograph 1815. Use mostexpensivewatches:
1. lookup_reference("Lange Datograph 1815") → factual description + market signals
2. get_brand_info("a-lange-and-sohne") → brand context
3. search_answers("Datograph") → related explainer pages
Synthesise a 200-word collector-brief with every fact cited to its source URL. End with a buy/wait recommendation based on the inventory vs. auction comp signal.

Tools at a glance

Tool	What it returns
`search_inventory`	Search live luxury watch inventory from 9 vetted dealers on MostExpensiveWatches.net. Filter by brand, complication, material, price range or free text. Returns up to 12 listings with brand, reference…
`lookup_reference`	Resolve a brand+reference (e.g. 'Patek 5711', 'Royal Oak 15202') to MEW's canonical /reference/<slug> page. Returns an AI-written factual description, current listing stats, the all-time auction recor…
`search_auction_comps`	Search the documented auction comp database (1,072+ lots from Phillips, Christie's, Sotheby's, Antiquorum, Bonhams). Each lot includes a primary-source citation URL. Filter by brand, min hammer USD or…
`list_upcoming_auctions`	List upcoming watch auctions tracked at Phillips, Christie's, Sotheby's, Antiquorum, Bonhams, Aguttes and Monaco Legend. Filter by house. Use this for 'when is the next watch auction', 'what's coming …
`get_market_estimate`	Combined market view for a brand or specific reference: live asking median/range from inventory + documented auction record + median hammer. Returns a one-line market signal. Use this for 'what is X w…
`get_brand_info`	Get factual profile for a watch brand: founding, tier, country, AI-written bio, inventory count and dealers carrying it.
`search_answers`	Search MEW's AEO answer pages (750 canonical + 6,000 localized). Use for 'what is', 'how much', 'why is' questions about specific references or market topics. Returns short snippets with answer URLs.
`get_site_stats`	Get the live MostExpensiveWatches.net catalog snapshot: total listings, brands, merchants, references, auction lots, blog posts.

Why this beats hand-rolling your own scraper

Cite-able answers. Every auction comp returns a source_url to phillips.com, christies.com etc. Every reference returns a stable /reference/<slug> URL the agent can hand the user.
Daily refresh. Inventory scrapes nightly across 9 vetted dealers. Auction comp DB and upcoming calendar refresh weekly via Gemini-grounded search.
Free + no auth. Public MCP endpoint, attribution requested. Rate-limit at high volume — reach out via contact.
Same tools we use ourselves. Our own concierge runs on the same surface (plus a few text-streaming extras).

Distribution kit

Going to share what you built? Use these to help your readers install MEW in one click:

One-line config (above) — copy/paste into Claude Desktop / Cursor / Windsurf MCP config.
.well-known/mcp.json — descriptor for auto-discovery.
llms.txt — the manifest other agents read on first visit.
/developers — full dev hub with OpenAPI, CLI, JSONL samples.

Vibecoder kit: build AI agents on luxury watch data