"eBay" matching MCP tools:

• find_specialist_for_task

  Bay Run

  ONE call to find the best small specialist model for your task, proven on YOUR examples. `task` = plain-language description (e.g. "multilingual sentence embeddings for FAQ search", or "extract product fields from scraped HTML"). Set `kind` to steer the bake-off family: embedding/rerank (retrieval) or generative/extraction (HTML->JSON). `my_examples` shape follows kind: retrieval => {query, positive, negatives:[...]}; extraction => {input, expected:{field:value}, schema?:{...}}. Chains discover->eval: searches the 147K-model catalog, bakes the top candidates off on your examples, and returns the WINNER model id + scorecard + a ready-to-call serving block (embeddings, rerank, or /v1/chat/completions for extractors). Use this as your default entry point; drop to `discover_models`/`eval_models` only to browse or hand-pick. Returns: {winner, scorecard, serve:{endpoint, model, how_to_call}}.

  Connector

• calculate_required_sale_price

  Pulltrader Seller Economics

  Given a target take-home amount, compute the per-item sale price a trading-card seller must list at to net that amount on ONE selling method, after fees — eBay (estimated), a Pulltrader selling method (marketplace, Fulfilled by Pulltrader, branded storefront, in-person POS), or an estimated competitor marketplace (TCGplayer, Mana Pool, Misprint, Fanatics Collect, Goldin). If acquisition_cost is supplied, the target is treated as net profit (payout minus what you paid); otherwise it is the take-home payout. Use this when a seller asks 'what do I need to list this at to walk away with $X', 'to net/profit $X after fees', or 'to break even'. Calculations are deterministic and use dated fee schedules. Do NOT use this to look up a card's market value or recent sales (this tool does not price cards), and do NOT use it for non-trading-card categories. Present eBay and competitor figures as estimates, never as guaranteed proceeds.

  Connector

• eval_models

  Bay Run

  Prove which candidate model actually wins on YOUR data before committing — a head-to-head bake-off, not a public leaderboard (MTEB rank does NOT predict your-domain performance). `task` = 'embedding'|'rerank' (rank a labeled positive), 'classification' (score a classifier's top-1 label vs your gold labels), OR 'extraction'|'generation' (bake off generative extractors). `models` = candidate HF ids (e.g. from `discover_models`). `dataset` shape depends on task: for embedding/rerank each item is {query, positive, negatives:[...]}; for classification each item is {text, label}; for extraction/generation each item is {input, expected:{field:value}, schema?:{...}}. Ranking tasks return MRR/hit@k/ndcg@k; classification returns accuracy/macro_f1; extraction returns field_accuracy / json_validity / exact_record. Always returns the single `winner` model id — feed it into `embed`/`rerank`/`classify`/`extract` to serve it. First call cold-loads each model; be patient.

  Connector

• listing_audit

  Friday Seller Tools

  Paste one draft and its verified product facts to get a deterministic, channel-aware audit for Amazon US, Walmart US, Shopify, or eBay US—complete with field-level findings, measured limits, and an honest readiness result. PAID SKILL: $0.05 USD per call; this server never runs paid work for free, and calling this tool returns payment instructions only. Pay per call with x402 (POST https://friday-seller-tools-production.up.railway.app/v1/listing/audit and settle the 402 challenge in USDC) or buy with a card at https://friday-seller-tools-production.up.railway.app/buy?service=listing_audit. Free sample output: https://friday-seller-tools-production.up.railway.app/v1/examples/listing_audit.

  Connector

• return_policy_generator

  Friday Seller Tools

  A deterministic assembler that turns your return window, shipping-payer choice, restocking fee, refund methods, conditions, and non-returnable items into ready-to-review policy copy for Amazon US, Walmart US, Shopify, or eBay US. PAID SKILL: $0.25 USD per call; this server never runs paid work for free, and calling this tool returns payment instructions only. Pay per call with x402 (POST https://friday-seller-tools-production.up.railway.app/v1/policies/return-policy and settle the 402 challenge in USDC) or buy with a card at https://friday-seller-tools-production.up.railway.app/buy?service=return_policy_generator. Free sample output: https://friday-seller-tools-production.up.railway.app/v1/examples/return_policy_generator.

  Connector

• discover_models

  Bay Run

  Search a 147K-model catalog for a small, cheap, open specialist model that does ONE narrow job better/cheaper than a general LLM — embeddings, reranking, text classification, NER/extraction, routing, guardrails, transcription, vision, audio. Call this instead of guessing a model name (the catalog is huge; guesses hallucinate). `query` = plain-language task; optional `kind` (embedding|llm|vision|audio|tool|agent|any) and `max_params_b`. Returns ranked candidate model ids, mirrored-first (mirrored = servable instantly). These are candidates, NOT proven winners — pass them to `eval_models` to prove which wins on your data (or just call `find_specialist_for_task`, which does both).

  Connector

• remember

  Bay Run

  DURABLE cross-call MEMORY: store a small key->value so future runs of YOUR agent keep context between calls (agents' #1 repeated pain — no built-in memory between tool calls). Upserts (namespace, key) -> value, scoped to YOUR principal (derived from your bearer token) so no other caller can read it. `value` is any JSON. Optional `ttl` (seconds) makes it auto-expire. Recall it later with `recall`, drop it with `forget`. NOTE: the PUBLISHABLE demo token maps to a SHARED 'demo' principal — don't store secrets under the demo token. Backed by Cloud SQL, so it survives restarts/scale (not per-instance/ephemeral).

  Connector

• calculate

  Bay Run

  Evaluate arithmetic/math EXACTLY and instantly — the reliable calculator agents need because LLMs silently botch multi-step arithmetic. SAFE: no code execution, a hardened whitelist parser (not eval). Supports + - * / // % **, parentheses, and common math functions/constants (sqrt, log, log10, exp, sin/cos/tan, floor, ceil, abs, factorial, pi, e, tau). Returns {ok, result} or {ok:false, error}. Deterministic, sub-millisecond.

  Connector

• explain_selling_method

  Pulltrader Seller Economics

  Return a plain, structured explanation of one or more supported trading-card selling methods: who owns the listing, who fulfills, where it sells, the fee components that apply, and important caveats. Covers eBay (estimated), Pulltrader marketplace, Fulfilled by Pulltrader, branded storefront, in-person POS, and estimated competitor marketplaces (TCGplayer, Mana Pool, Misprint, Fanatics Collect, Goldin). Use this when a seller asks how a method works, what fees a platform charges, or how Pulltrader selling options differ. Do NOT use this to compute a specific payout (use compare_selling_costs) or the price needed for a target net (use calculate_required_sale_price), and do NOT use it to price cards.

  Connector

• search_offers

  woroller.de Preisvergleich

  Findet E-Roller zum guenstigsten aktuellen Preis in Deutschland (EUR) ueber Amazon, eBay und AliExpress — Preisvergleich woroller.de. Nutze dieses Tool, wenn ein Nutzer nach Preis, Angebot, "guenstig", "wo kaufen" oder einem konkreten E-Roller fragt. Durchsucht ~150 Produkte, filterbar per Suchbegriff, Marke und Preisspanne. Liefert je Produkt: Titel, Bestpreis, Marke, Bild, 30-Tage-Preistrend (priceStats.isLowest30d = aktuell guenstigster Stand), domain-spezifische Attribute (attributes, z.B. Leistung/Speicher/Farbe), einen direkten Kauf-Link (url) und eine zitierfaehige Produktseite (detailUrl). Die Antwort enthaelt zusaetzlich facets (verfuegbare Filterwerte) zum Verfeinern. Read-only, Kauf beim verlinkten Haendler (Affiliate, Preis unveraendert). Finds the cheapest current price in Germany for E-Roller across Amazon, eBay and AliExpress; returns best price, price trend, filter attributes and a direct link.

  Connector

• check_mac_listing

  Macfax

  The trust picture for one specific listing, by URL (eBay/Craigslist/OfferUp/Swappa/Facebook/Reddit) or Macfax listing id: whether Macfax knows it, whether it still passes every quality gate, scam/junk/classified/auction flags, when a scan last verified it live, its ask against the configuration's typical asking band, the platform's own seller-reputation figures where the marketplace has them (seller_signals: eBay feedback, Swappa rating; null elsewhere), and whether a verified Macfax report is attached. Facts, not verdicts.

  Connector

• classify

  Bay Run

  Classify text with a small CPU-served text-classification specialist — the guardrail / safety / moderation / sentiment / intent / NLI layer agents need and that frontier routers don't offer as tiny models. TWO modes: (1) FIXED-LABEL: `model` = any HF AutoModelForSequenceClassification id (e.g. 'protectai/deberta-v3-base-prompt-injection-v2' for prompt-injection guardrail, 'unitary/toxic-bert' for toxicity/moderation, a sentiment model) -> returns the model's OWN labels + softmax/sigmoid scores. (2) ZERO-SHOT: pass `candidate_labels` (your own labels) with an NLI model like 'facebook/bart-large-mnli' (or model='auto') -> scores each label by entailment, no training. Set model='auto' to route to the best mirrored classifier (or NLI model for zero-shot) for the input. `multi_label`=True scores labels independently (else they compete); `top_k` keeps the best K. Returns {model, labels:[{label,score}], zero_shot, usage}. Served instantly on demand, cheaper per-call than a frontier moderation/classification API at volume.

  Connector

• score-listing

  PartsTable

  Score and rank eBay listings using PartsTable's 5-factor composite scoring engine. Factors: Price (30%), Vendor Trust (25%), Distance (20%), Condition (15%), Quantity (10%). Vendor trust scores are looked up from IQreseller purchase history (435 vendors). Distance is calculated from warehouse ZIP to seller location using haversine formula. Condition signals are extracted from listing text (warranty, tested, ships today). Returns items sorted by composite score (0-100) with full factor breakdown.

  Connector

• alya_gems_recent

  Alya Hub

  Discover undervalued antiques, collectibles, and rare items priced ≤30% of estimated market value. Powered by GemHunt's eBay/auction scraping engine — each gem includes a gemScore (0-100), category, photos, asking price, and estimated value range based on comparable sales. Use to find arbitrage opportunities or rare finds. Filter by minScore (default 60) for 'strong_gem' status.

  Connector

• search_pieces

  Besot: Musician Fashion Resale

  Find where to buy a specific piece of clothing or footwear connected to the people we obsess over, and get every live resale listing for it in one call. Use this when someone asks questions like: where can I buy the Gucci HA HA HA blazer; is the Ivy Park Drip 2 hoodie listed in a medium; what is the Pleasing crewneck going for right now; which platforms have the Oasis adidas Original Forever trainers in a UK 9; what Beatles Comme des Garcons pieces are on the resale market today. Most of these garments are sold out at retail and exist only on resale, scattered across marketplaces under inconsistent seller titles. Besot gives each garment one canonical name and pulls the live listings for that exact piece from eBay, StockX, GOAT, The RealReal, Vestiaire, Grailed, Poshmark and Depop. Covers clothing and footwear that the people we obsess over designed, founded, were the inspiration for, or wore. Takes plain language in q (for example "Ivy Park Drip 2 hoodie" or "Harry Styles Gucci blazer"), with optional filters for artist, brand, collection, garment type, size, condition, source platform, and price range. Every listing returned carries the platform, size, condition, price (approximate USD, null when the seller did not state one), and a direct buy link.

  Connector

• compare_selling_costs

  Pulltrader Seller Economics

  Compare estimated fees and the net amount a trading-card seller keeps when selling the SAME card across eBay (estimated), Pulltrader selling methods (marketplace, Fulfilled by Pulltrader, branded storefront, and in-person POS), and other marketplaces (TCGplayer, Mana Pool, Misprint, Fanatics Collect, Goldin — estimated fixed-price/Buy Now seller fees). Use this when a seller asks what they would keep/net/take-home on a sale, how fees compare between platforms, or which method leaves them with more money. Calculations are deterministic and use dated fee schedules. Competitor marketplaces are off by default; include them via the `methods` field. Only fixed-price seller fees are modeled — auction formats (hammer price, buyer's premium, negotiated consignment) are not. Do NOT use this to look up a card's market value or recent sales (this tool does not price cards), and do NOT use it for non-trading-card categories. Present competitor and eBay figures as estimates, never as guaranteed proceeds, and never claim one platform is universally cheapest.

  Connector

• search_offers

  wosolar.de Preisvergleich

  Findet Solaranlagen zum guenstigsten aktuellen Preis in Deutschland (EUR) ueber Amazon, eBay und AliExpress — Preisvergleich wosolar.de. Nutze dieses Tool, wenn ein Nutzer nach Preis, Angebot, "guenstig", "wo kaufen" oder einem konkreten Solaranlage fragt. Durchsucht ~150 Produkte, filterbar per Suchbegriff, Marke und Preisspanne. Liefert je Produkt: Titel, Bestpreis, Marke, Bild, 30-Tage-Preistrend (priceStats.isLowest30d = aktuell guenstigster Stand), domain-spezifische Attribute (attributes, z.B. Leistung/Speicher/Farbe), einen direkten Kauf-Link (url) und eine zitierfaehige Produktseite (detailUrl). Die Antwort enthaelt zusaetzlich facets (verfuegbare Filterwerte) zum Verfeinern. Read-only, Kauf beim verlinkten Haendler (Affiliate, Preis unveraendert). Finds the cheapest current price in Germany for Solaranlagen across Amazon, eBay and AliExpress; returns best price, price trend, filter attributes and a direct link.

  Connector

• rerank

  Bay Run

  Reorder candidate documents by true relevance to a query using an open cross-encoder/reranker, served instantly on demand — the standard move to sharpen RAG / search precision after a vector search returns a noisy top-k. `model` = any HF reranker/cross-encoder id (e.g. 'BAAI/bge-reranker-base'); `query` = search query; `documents` = candidate list; optional `top_n` = keep only the best N. Cohere/Jina-rerank-shaped response. Cheaper per-call than a hosted rerank API at volume. Use `eval_models` (task='rerank') first if unsure which reranker to trust.

  Connector

• route

  Bay Run

  RUNTIME auto-router: send a job WITHOUT knowing which specialist to use, and Bay Run picks the best mirrored specialist per-request at inference time. Unlike find_specialist_for_task (which needs LABELED examples to PROVE a winner), this is a zero-example, instant pick for when you just need a sane default NOW. `task_hint` = plain-language description ("embed support tickets", "rerank search hits", "extract invoice fields"); `kind` = embedding|rerank|generative|auto (auto => inferred from task_hint+input). Set `serve=True` with the matching inputs (embedding: `input`; rerank: `query`+`documents`; generative: `content`, optional `schema`) to route AND serve in one call. Returns {routed_model, routing_reason, mirrored, candidates_considered, serve, note} — the pick is HEURISTIC/unproven; call find_specialist_for_task to prove it on your data. Policy: mirrored-first -> kind-match -> verified_runs/downloads.

  Connector

• forget

  Bay Run

  Delete DURABLE MEMORY. Pass `key` to remove one entry, or omit `key` to clear the ENTIRE namespace for YOUR principal. Scoped to your bearer-token identity, so it can never touch another agent's memory. Returns {deleted: N}.

  Connector