Dealectics deal verdict
Server Details
Ask before you buy: is this price a good deal, from buyer-approved offers.
- Status
- Healthy
- Uptime
- 100.0% over the last 24 hours
- Last Tested
- Transport
- Streamable HTTP · MCP 2025-11-25
- URL
- ChatGPT
- –
- Claude
- –
TDQS
Scored across 5 tools
The watch lifecycle tools create_watch, get_watch, and stop_watch are clearly separated by action. price_verdict and search_deals are explicitly scoped: one is for a specific product at a known price, the other for vague or category searches. An agent can choose reliably between all five.
create_watch, get_watch, stop_watch, and search_deals follow a clear verb_noun pattern. price_verdict is the only deviation from that pattern, but it remains readable. This is a minor inconsistency rather than a systemic problem.
Five tools is well scoped for a deal-verdict and watch-management server. Each tool covers a distinct action and no obvious redundant tool is included. The set is neither thin nor bloated.
The surface covers watch creation, status, stopping, specific price verdicts, and deal search. A minor gap is the lack of a list_watches tool or another way to discover existing watches without already knowing their IDs. Confirmation being user-driven is intentionally out of scope, so agents can still work around the gap.
Available Tools
5 toolscreate_watchWatch for a deal and e-mail the buyerAInspect
Use when the person wants to be told when a good deal appears, and has agreed to receive e-mail. Sends them a confirmation link; nothing is watched until they open it, and you cannot confirm for them.
Inputs: need (required), email (required), user_consented (required), max_price.
get_watchRead a watchARead-onlyIdempotentInspect
State of a watch created with create_watch: awaiting confirmation, active or stopped, and how many alerts were sent.
Inputs: watch_id (required).
price_verdictIs this price a good deal?ARead-onlyIdempotentInspect
Use before buying or recommending a specific product at a known price in euros. Give brand and model (and the merchant page if you have it) with the price. Returns 'bon', 'moyen', 'mauvais' or 'inconnu', the price range buyers approved, the usual price they report, how many observations support it and how recent they are, and what stays unknown. Each sampled observation gives its shop, the shop's market and whether it is cross-border for the country asked. Prices of used or refurbished products are left out and counted apart; variant is never compared. In approved_price_range, low, typical and high are quartiles; lowest and highest are the extremes. 'inconnu' means the evidence is too thin: say so rather than guess. Not for categories ('a laptop'): use search_deals.
Inputs: price (required), url, country, product, +1 more.
search_dealsFind current deals for a needARead-onlyIdempotentInspect
Use when someone wants to buy something, precise or vague ('noise-cancelling headphones under 200 €', 'a gift for a DIY dad'). Returns at most five offers seen in the last three days in France or Germany, each with price, merchant link when known, separate relevance, deal-quality and purchase-feasibility signals, and its unknowns; or 'no_offer' with the reason. max_price, ship_to and exclude are hard limits. Offers marked 'lead' are real but unconfirmed: check the merchant page before presenting them as available. Each offer has a merchant_link status: 'resolved' (merchant_url is the product page), 'pending' (being looked up: repeat the same search after retry_after_seconds, once), 'abandoned' with a reason (no lookup is under way: use merchant_search_url) or 'unknown'. merchant_market is the country of the shop and cross_border is true when it differs from ship_to: delivery there is then unverified. delivery gives the shipping cost the listing stated, not a quote for an address. id identifies the same offer across calls. The merchant's live price is never rechecked: purchase_feasibility stays below 1.0 and every limit on it is named in unknowns. 'no_offer' comes with the limits applied and why recent candidates were left out.
Inputs: query (required), exclude, ship_to, currency, +2 more.
stop_watchStop a watchADestructiveIdempotentInspect
Stop a watch created with create_watch. Safe to repeat.
Inputs: watch_id (required).
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Related MCP Connectors
Negotiates with shops and providers for prices not on their websites, and returns every offer.
Does this retailer take offers, and does it price match? Verified answers, checked by a person.
Evidence-backed pre-purchase checks for offer terms, seller trust and policy compliance.
Ask Deal.ai's shopping agent, compare every US store's price, check a deal and watch prices.
Related MCP Servers
- AlicenseNot gradedqualityBmaintenanceLets shopping agents turn a single purchase into a live auction where merchants bid, verifies every bid against the merchant's real product page and checkout, and ranks the surviving offers on merit rather than fees. It enforces spending limits with text-based human approval, then settles approved purchases in a double-entry ledger.1MIT
- AlicenseNot gradedqualityBmaintenanceEnables users to organize purchase criteria, search indexed marketplace offers, and compare total costs including shipping, taxes, fees, and confirmed coupons to identify the best value.MIT
- AlicenseNot gradedqualityCmaintenanceEnables natural-language price negotiation by turning purchase requests into ranked cheapest-first estimates and drafting vendor messages.Apache 2.0
- AlicenseAqualityCmaintenanceEnables searching Facebook Marketplace listings, sweeping query variants and sort orders to group comparable items like-for-like, compute price stats and deal scores, and flag scam or low-quality listings. It also drafts opening offers, saves watches with target prices, and reports new listings, price drops, and local price history, with Hebrew/English and Israeli ₪ support.8MIT
Glama MCP Gateway
Add one secure layer between your agents and this server.