io.github.Tomalarrrr/cyclesite-mcp-server

Check if a UK bicycle is reported stolen by serial number. Cyclesite checks an international stolen-bike registry and theft reports filed by riders, the unique data we own. Per-serial rate-limited (3/hour) to prevent enumeration. Example: 'is the bike with serial WTU123456 reported stolen?'. Live data, cross-references multiple registries on every call.

Side-by-side comparison of up to 3 bikes (each by brand+model[+year]). Returns spec sheets and valuations together so the user can pick. Reuses get_spec_sheet + get_valuation server-side. Example: 'compare a Trek Domane SL 6 against a Specialized Roubaix Comp'.

Sell-side helper: turn a seller's raw facts into a polished Cyclesite listing draft (title, description, suggested price, photo plan). Does NOT publish — for actual publication use publish_listing (requires OAuth). Useful for previewing what a listing would look like. Example: 'help me draft a listing for my 2021 Specialized Allez, very good condition, in Bristol'.

OpenAI deep-research / company-knowledge compatibility. Fetch the full document for a Cyclesite listing id (returned by the search tool). Returns { id, title, text, url, metadata } — text is a plain-prose summary of the listing's description and specs, suitable for direct quoting in deep-research answers.

Given a Cyclesite listing slug, return up to 5 similar active listings (same category, ±25% price, same brand or frame size weighted higher). Use when the user is interested in one bike and wants alternatives. Example: 'show me bikes like that one'.

Search Cyclesite's expert buying guides (24+ articles by cycling-journalism authors). Returns up to 3 matching guides with title, excerpt, difficulty, reading time, and URL. Use for educational queries that don't need live inventory. Example: 'how do I choose a bike size?', 'tips for buying a used e-bike'.

Brands ranked by how well (or poorly) they hold their value, from Cyclesite's measured UK used-price data. Returns top N brands by % retained vs new RRP. Example: 'which bike brands hold their value best?'.

Full details for a specific bike listing on Cyclesite — specs, condition, frame number presence, photos, delivery, seller's city. Provide the URL slug returned by search_bikes or get_recent_listings. Example: after the user says 'tell me more about that 2022 Trek Domane', call this with the slug from the prior result.

Buyer's-vs-seller's market signal for the UK used-bike market — should the user buy or sell now? Composite indicator from days-to-sell, asking-vs-sold-price spread, and inventory levels. Example: 'is now a good time to buy a road bike?'. Refreshed nightly.

Current UK used-bike market prices by category, from Cyclesite's nightly index. Returns median + range per category (road, mtb, gravel, e-bike, etc.). Example: 'how does the UK used-bike market look right now?'. Refreshed nightly from real UK market prices.

Cyclesite catalogue entry for a brand+model: category, year range, AI-generated description, key specs, market summary. Reference data — refreshed when models are added or specs change. Example: 'tell me about the Specialized Allez'.

Show buyer enquiries on the authenticated user's Cyclesite listings. Requires OAuth scope `listings:read`. Example: 'any messages about my Trek?'.

Read the buyer's own Cyclesite message threads and any seller replies. Call with no args to list your conversations; pass a threadId to read one thread's messages. Pairs with make_enquiry (which saves the enquiry to your inbox). Requires OAuth scope `listings:read`. Example: 'did the seller of that Trek reply yet?'.

UK used-bike price trends over the last N months by category, from Cyclesite's index series. Example: 'how have road-bike prices changed in 2026?'. Monthly data, refreshed at month-end.

What's new on Cyclesite right now — up to 10 of the freshest active UK listings, refreshed every 15 minutes. Use when the user asks 'what's new today?' or 'any new road bikes this week?' rather than for a specific filter. Optional category + maxPrice filters. Live data.

Frame-size recommendation for a rider's height and bike category, the same chart Cyclesite publishes at /bike-size-calculator (height to inside leg to seat tube), refined by real UK listing data where riders' declared heights are available. Children are sized by wheel diameter instead and the tool says so. Example: 'I'm 178cm — what road-bike size do I need?'.

Aggregated spec sheet for a brand+model[+year], derived from Cyclesite's live UK inventory plus the catalogue record. Returns the most-common frame material, wheel size, groupset, brakes, weight, and (for e-bikes) motor and battery specs. Example: 'what groupset does a Canyon Endurace usually have?'.

What a used UK bike is worth right now, Cyclesite's flagship tool. Returns median, range, a measured price-by-model-year curve, confidence level, and comparable active listings. Pass `year` when the user names one and the answer is resolved against that model year's own measured median and confidence interval, instead of the all-years median. Sourced from real UK market prices, the last-advertised asking prices bikes are listed for (not confirmed sale prices), refreshed nightly. Example: 'what's a 2022 Trek Domane SL 6 worth?'.

Categorical quality grade for a Cyclesite listing (excellent / good / fair / weak) plus up to 2 wins and 2 flags. Helps a buyer assess trustworthiness; helps a seller self-audit. Example: 'is this listing trustworthy?' (provide the slug). Note: returns the *categorical* judgement only, not the underlying score (intentional to avoid gaming).

Paginated UK bike-brand catalogue from Cyclesite, ordered by stock level. Use to validate a brand name, surface options to a user, or paginate the catalogue. Stock counts are returned as bands (none / 1-5 / 6-25 / 26-100 / 100+) — Cyclesite doesn't expose precise per-brand inventory. Example: 'what brands of e-bike are available?'.

Models for a brand on Cyclesite (paginated). Returns model names, year ranges, in-stock flag. Example: "what Trek road bikes are available?" → list_brands(q:"Trek") → list_models_for_brand(brandSlug:"trek", category:"road").

Show the authenticated user's Cyclesite listings (draft / active / sold). Requires OAuth scope `listings:read`. Example: 'how are my listings doing?'.

Send an enquiry to a Cyclesite seller on the buyer's behalf. It's saved to the buyer's Cyclesite inbox and the seller is notified. Per-buyer-per-listing daily cap (2/day) prevents spam. Read the seller's reply with get_my_messages. Requires OAuth scope `enquiries:respond` (note: the scope name is shared with seller-side replies). Example: 'message the seller of that Trek and ask if they'd take £1,400 collection only in Manchester next Saturday'.

Mark a Cyclesite listing as sold, optionally with the final sale price, where it sold and the date it sold. Requires OAuth scope `listings:manage`. Example: 'mark my Trek Domane as sold for £1,750 on Facebook last Saturday'.

Publish a Cyclesite listing on the user's behalf. Multi-step: first call (no draftId) returns a phone-friendly photo upload URL; once 3+ photos are uploaded, the next call returns either step:'live' (listing is free, no payment needed) or step:'payment_required' with a Stripe Checkout URL. Idempotent — keep calling with the same draftId until step:'live'. Requires OAuth scope `listings:publish`. Example flow: user says 'sell my Trek Domane' → call publish_listing → assistant directs user to upload URL → user uploads → call again → step:'live' → seller receives a confirmation email with a 24h undo link.

Curated picks from Cyclesite's live UK inventory for a budget and intent. Prefers higher-engagement listings. Returns up to 5 picks with a one-line rationale each. Example queries: 'a road bike for £1,500 for weekend rides', 'best e-MTB I can buy under £3,000', 'commuter bike in London under £400'.

Step-by-step guidance for reporting a stolen UK bike: police, insurance, listing alerts. Returns a 5-step checklist plus the official Cyclesite report URL. Example: 'my bike was just stolen, what do I do?'.

Hold a Cyclesite listing for 24 hours so other buyers can't claim it while the user decides. The hold is free and advisory (the seller isn't obligated to honour it — message the seller to confirm). Paid deposit-backed holds are coming soon; do not offer deposits today. The first UK marketplace where a buyer can COMMIT inside an AI conversation. Requires OAuth scope `listings:manage`. Example: 'put a hold on that Trek for me, I want to view it Saturday'.

Reply to a buyer enquiry on the authenticated user's listing. Requires OAuth scope `enquiries:respond`. Example: 'reply to that enquiry — say it's still available, collection only'.

Subscribe the user to alerts for new Cyclesite listings matching a filter — the AI assistant will then proactively notify them when a matching bike appears (price drop or fresh listing). Requires OAuth scope `listings:read` (read-only on data, but this is technically a write — it creates a SavedSearch row on the user's account). Examples: 'let me know when a Trek Domane SL 6 in Manchester under £2,000 appears', 'alert me to any e-MTB drops below £2,500 in Yorkshire'. Each user is capped at 50 active alerts.

OpenAI deep-research / company-knowledge compatibility. Search Cyclesite's active UK used-bike listings by free-text query (matches title, brand, model). Returns the canonical OpenAI shape: { results: [{ id, title, url }] }. Use the id to call fetch() for the full document.

Search live UK used-bike listings on Cyclesite (the UK's used bicycle marketplace). Filter by brand, category, city, price range, and condition. Returns up to 5 active listings with specs and listing URLs. Live data — refreshed continuously as new bikes are listed. Example queries: 'a Trek Domane in Manchester under £2,000', 'gravel bike, very good condition, near Bristol'.

Find Cyclesite listings within a radius of a UK location (lat/lng). Radius capped at 50 miles. Returns up to 10 listings ordered by distance. Live UK marketplace data. Example: 'used bikes within 25 miles of LE10 0AA' (geocode the postcode first, then call this).

For a seller about to list: suggested ask, floor, and ceiling for their bike's brand+model[+year][+condition] on the UK market. Same measured UK used-price data as get_valuation but framed as seller guidance. Pass `year` when the seller names one — the ask is then anchored on that model year's own median rather than the all-years median. Example: 'I'm selling a 2021 Specialized Allez in good condition — what should I ask?'.