MCP App Store
by KeenEthics
CarEdge icon

CarEdge MCP Server for Claude

By CarEdge

Overview

CarEdge connects your AI assistant to cars for sale at dealerships across the United States, and to CarEdge partner dealers for itemized quotes and contact requests. What you can do: - Search cars in stock by make, model, year, body style, budget, new or used, and distance from your ZIP code. - See one car's details: specs, features, photos, mileage, colors, and any incentives the dealer lists. - Get an itemized out-the-door price for your ZIP code at a CarEdge partner dealer: the selling price, dealer fees, and, where the dealer quotes them, sales tax, title and registration. Some dealers quote before tax; the quote says so. A quote is marked as an estimate unless the dealer has confirmed its fees. At other dealers you get the listing price, a CarEdge out-the-door estimate and a link to the car on caredge.com. - Find dealers near you, with phone (where the listing has one), distance and CarEdge transparency grade. - Decode a VIN, check NHTSA recalls, and estimate a trade-in value or a monthly payment. - Ask a CarEdge partner dealer to contact you about a car. Your assistant asks for your name and how to reach you, shows the dealer's consent wording where it has one, and asks whether the dealer may contact you. Only then is your request sent to that dealer, and it says which assistant you used. How it works: searching, pricing and checking dealers need no account. You sign in with a CarEdge account, or create one, only when you ask a dealer to contact you. Searching and pricing never send your details to a dealer. Contact details go to a dealer only when you ask to be contacted about a specific car. Limits: itemized quotes and contact requests are for CarEdge partner dealers only. Prices and stock come from the dealer's own listings and can change; the dealer confirms the final price. CarEdge does not sell cars, take payments or arrange financing.

How to connect

  1. Claude: Open the CarEdge listing in the Claude connectors directory and choose Connect. Claude listing CarEdge exposes 14 tools here.

Tools

calculate_finance_payment

Claude
Use this when the shopper asks what a car would cost per month, e.g. vehicle_price: 31250, down_payment: 3000, state: "OK". It needs a real price, from the shopper or a tool. Do not use it for a dealer's out-the-door price (that is quote_out_the_door). It works out a monthly loan payment from the price, down payment, trade-in value, term and APR. With no apr it shows example payments for excellent, good, fair and poor credit, and rates_note says those APRs are examples. Amounts are in dollars. With state it adds a state-level sales tax estimate that follows the state's trade-in credit rule (full credit in 47 of 51 jurisdictions, none in CA, DC, OK and VA), in tax_estimate, taxable_amount, trade_in_credit_policy and trade_in_credit_cap. With no state, or one it does not know, it adds no tax: tax_estimate is 0, unknown_state is true and tax_note says the payment excludes tax. caveat, when set, states the limits of the tax figure. County, city and district tax are not included. When the price is already an out-the-door figure with tax and fees in, such as a dealer's quote, price_includes_taxes_fees: true adds no tax, and the loan is the price less the down payment and the trade-in.

check_vehicle_recalls

Claude
Use this when the shopper asks whether a model has open safety recalls, e.g. year: 2022, make: "Ford", model: "F-150". Do not use it for a car's accident or title history. It reads NHTSA's open recall campaigns for that year, make and model, each with the affected component and the remedy. At most 10 recalls are listed; total_recalls is NHTSA's full count and truncated says when there are more. upstream_unavailable means NHTSA did not answer, so the recall check was not done; it does not mean the car is clear.

dealer_info

Claude
Use this when the shopper asks about one dealership rather than a car: which brands it sells new, how many cars it has, its doc fee, whether its prices include tax, or how transparent it is on price, e.g. rooftop_slug: "jim-glover-chevrolet". It does not know opening hours; the answer has the phone and website. Do not use it to list stores (that is stores_list) or to find cars (that is inventory_search). When the connector covers one store, rooftop_slug may be left out; otherwise pass one from stores_list, or the answer is invalid_params. A rooftop_slug of the form dealer-<id> from stores_list describes a dealer in CarEdge's marketplace that is not a partner, from its listings: participating: false, no fees, and its DTI grade where one exists. brands_new is read from the new cars in stock, so a franchise with no new car on the lot right now is not listed. doc_fee and other_fees come from the store's fee schedule, and are null when the store takes no quotes. doc_fee is also null when the store names that fee differently; it then appears in other_fees under the store's own label. quotes_include_taxes: false means its quotes stop before tax, title and registration. binding_fees: false means the fee schedule is not confirmed yet, so every quote is an estimate. dti_grade (A to F) and dti_score (0 to 100) are CarEdge's Dealer Transparency Index, from real out-the-door quotes; both are null when the store has no grade yet.

decode_vin

Claude
Use this when the shopper pastes or gives a VIN and you need to know what car it is (year, make, model, trim, engine), e.g. vin: "1GNSKRKD2PR339514". Do not use it for a car for sale that the shopper picked from a search (that is vehicle_get). It decodes the VIN with NHTSA's free vPIC database into year, make, model, trim, body type, fuel type, drivetrain and engine. It suits a VIN pasted with nothing else, or a VIN with no year, make or model known. When the VIN is a car on this dealer's own lot, the answer also carries in_dealer_stock, store_name and dealer_stock_note: it is a car for sale here, not the shopper's own, and vehicle_get has its listing.

estimate_trade_in

Claude
Use this when the shopper wants to know roughly what their own car is worth as a trade-in, e.g. year: 2020, make: "Honda", model: "Accord", trim: "EX-L", mileage: 45000 (add vin when they give it). Do not use it to value a car a dealer is selling (that is vehicle_get). It values the car the shopper wants to trade in or sell, for example to work out the net cost of a new car. With a vin it returns trade-in values for that exact trim and condition; without one it estimates from retail listings, which is less exact. A vin is decoded first, and the year, make and model it decodes as are the car valued, in vin_decoded_as. source says which (vehicle_database or comparable_listings), and guidance is a note on how firm the figure is and how it was reached. On the estimate, market_data.trim_matched: false means the listings are of another trim, so the figure is based on similar cars, not this exact one. Without trim and mileage the comparison covers every trim and mileage of the model. Money is in cents, as numbers. not_found means no comparable listing with a price was found for that year, make and model, so there is no figure.

inventory_facets

Claude
Use this when the shopper asks what the dealer has rather than for one car: which makes, models or trims are in stock, which colors, how many trucks, or the cheapest and dearest price, e.g. body_type: "pickup". It counts every matching car, so answer from it instead of counting search results. Do not use it to show cars (that is inventory_search). It counts every car that matches, not one page: how many in all, and how many of each make, model, trim, body type, drivetrain, fuel and color, plus the lowest and highest year, listed price before fees, and miles. Use it for whole-lot questions ("what trucks do you have?", "which colors does the Tahoe come in?", "do you have anything electric?"), or before searching when you do not know what to filter on. It lists no cars: inventory_search does that, with the make, model or trim the shopper picks. It takes the same filters as inventory_search: when these stores span the whole country, pass query, make or near, or the answer is too_broad. With near and no rooftop_slug, it counts every dealer in CarEdge's marketplace within the radius, the same cars inventory_search finds there. Each list is most common first and holds at most 20 values. The filters only inventory_search has, such as features or dti_grade, are not taken here. filter_values holds, under each inventory_search argument's name, the values that argument can take among these cars, with counts, in the spelling it accepts. days_on_market is the fewest and most days these cars have been listed. price_before_fees is the range of listed prices before the store's fees, in dollars. The cars themselves show price_with_fees, which is higher by the store's fees, so it is not any one car's price. The cheapest car's price_with_fees comes from inventory_search with sort: "price_asc" and limit: 1. freshness says how old each store's data is, and inventory_unavailable means CarEdge cannot see the lot right now, not that it is empty.

lead_prepare

Claude
Use this when the shopper asks to be contacted by the dealer about one car, for example to check it is available, book a test drive or buy it, e.g. channels: ["email", "sms"]. Do not use it to look up a car or a price, or before the shopper has asked to be contacted. Nothing is sent yet: this is step one, and lead_submit is step two. It needs the shopper's first name and at least one of an email address or a phone number; a last name is optional. channels is what the shopper agreed to be contacted on, and the dealer is never asked to use more than what is passed here. It returns disclosure_text, the store's consent wording, or CarEdge's when the store has none, and a ticket that is good for fifteen minutes. The shopper has not agreed yet. lead_submit comes in a later turn, once the shopper has seen that exact text and replied agreeing to it (a clear yes, or their details), with the text unchanged and shopper_consented: true. Approving a tool call in the app is not agreement. An unused ticket expires by itself and creates no lead. An empty disclosure_text means no consent wording is set (consent_wording_note says so); the shopper's yes to being contacted by the store is still needed. A store with no consent wording of its own (participating: true and disclosure_text: null in vehicle_get) takes CarEdge's wording, which covers email only: it accepts an email address and channels: ["email"], with no phone number. rooftop_slug and buyer_zip decide which store receives the lead when the same car sits at more than one; routing_reason says which rule applied. quote_id from quote_out_the_door records on the lead the figure the shopper was actually shown. A shopper may have three leads prepared a day. Past that, if this shopper's lead for this car was already sent in the last 24 hours, the answer is that lead, in lead_submit's shape with already_sent: true and no ticket: the request is already with the dealer and there is nothing to submit. Otherwise the answer is rate_limited. comments may be at most 500 characters. A conversation follow-up is the exception: a lead sent after the chat, with the chat in comments, that adds to a lead sent for this shopper and car in the last 24 hours. Set followup_type: "conversation" and adds_to to that lead's lead_id, never a follow-up's. customer and channels may then be left out and are that lead's; if sent, they may add no email, phone or channel that lead lacks. Its consent carries over, so disclosure_text comes back empty and nobody is asked again; if the store has added consent wording since, the answer is consent_required. Its comments may be longer, up to the length the server allows; longer is refused, never cut. It counts toward the limits above, and only stores switched on for it take one. A trade-in follow-up (followup_type: "trade_in", or none) keeps the rules of a first lead and names adds_to at lead_submit. A car outside the stores this connector covers answers not_found. A store CarEdge has no lead arrangement with answers quote_unavailable: that car can be described, but no lead can be handed over for it. A car at a dealer that is not a CarEdge partner answers quote_unavailable: CarEdge cannot send that dealer a lead, nothing is saved or sent, and the message gives the car's caredge.com page and, when it can be resolved, the dealer's.

lead_submit

Claude
Use this to send the shopper's request to the dealer once they have agreed to the consent wording, with the ticket from lead_prepare. It reaches a real dealership. Do not use it without a ticket from lead_prepare. It sends a real lead. It is refused with consent_required unless disclosure_text is exactly the text lead_prepare returned, customer is the same shopper, and shopper_consented is true, because CarEdge must be able to show that this shopper agreed to these words. Agreement is the shopper's own reply in a later turn, after seeing the text: a clear yes, or their details. Approving a tool call in the app is not agreement. When lead_prepare returned an empty disclosure_text, disclosure_text may be left out, and shopper_consented: true is still needed: the shopper's yes to being contacted by the store. channels may be left out, and then the dealer is asked to use exactly the channels chosen at lead_prepare. It may name fewer, never more. A ticket makes one lead. The exact same call sent again answers with the same lead_id and the lead's current status, and nothing is sent twice. A different request on a ticket already used answers ticket_used. After fifteen minutes the ticket answers ticket_expired, and a new lead_prepare, with the consent wording shown again, is needed. One shopper and one car make one lead a day. When this shopper's lead for this car was already sent in the last 24 hours, from any chat, the answer is that lead, with its own lead_id and accepted_at, and already_sent: true: the request is already with the dealer and nothing new reaches the store. Something the shopper gave after that lead, such as a trade-in, goes in a new lead whose comments say what is new, with adds_to set to the earlier lead_id; it reaches the store as its own lead. trade_in holds the shopper's own car, and the dealer's CRM files it as the trade-in vehicle. Each store takes at most 25 leads a day from AI assistants. Past that the answer is rate_limited, and the store can still be called directly. On a ticket lead_prepare gave for a conversation follow-up, the earlier lead's shopper and consent carry over: customer, disclosure_text and shopper_consented may be left out, and adds_to is already in the ticket. comments may be as long as lead_prepare allowed; on every other lead they are at most 500 characters. status: "accepted" means the lead is CarEdge's and safe. The email to the dealer's CRM goes out straight afterwards and is retried on its own if it fails, so a delivery problem is never the shopper's to solve. A dealer that is not a CarEdge partner never gets a ticket, because lead_prepare answers quote_unavailable for its cars, so no lead can reach such a dealer.

lookup_dealer_transparency

Claude
Use this when the shopper asks whether a dealer is legit or can be trusted on price, how honest a dealer is about fees and add-ons, what its doc fee usually is, or how its fees compare, e.g. dealer_name: "Towne Hyundai", state: "NY". The grade comes from real out-the-door quotes from that dealer, not from reviews. It does not check licences, reviews or complaints. Do not use it to find stores (that is stores_list). It returns the dealer's CarEdge Dealer Transparency Index grade (A to F) and score, the number of out-the-door quotes the grade is based on (quote_count), its average and median doc fee, add-on behavior and price accuracy, and doc_fee_benchmarks: the state and national average doc fee from CarEdge's State of Dealer Fees report. With a VIN, the grade is for the store selling that car, which matters for dealer groups with several locations. The state tells apart dealers with the same name. When a name matches dealers in more than one state and no state or VIN is given, dealer is null and matches lists them (name, city, state, grade, quote count); the same call with one of those states returns that dealer. dealer is null, with a message, when CarEdge holds no grade for that dealer; similar_names then lists graded dealers sharing the name's other words, which may be different dealers. A dealer graded F because it never sent a written price has grade_reason saying so, quote_count 0 and null scores and fees. Any other dealer with no quotes has a null grade. A fee or score CarEdge has not recorded is null, not zero.

quote_out_the_door

Claude
Use this when the shopper asks what one car will really cost them, its total or out-the-door price, or how much the dealer's fees add: the price with that dealer's own fees and, where the store quotes them, sales tax, title and registration for the shopper's ZIP, e.g. vin: "3GNKBKRS6PS118420", buyer_location: {"zip": "74133"}. Do not use it for monthly payments, loan or lease terms, or what a trade-in is worth. The ZIP where the shopper will register the car sets the sales tax, the registration and the title fee, and the same car costs a different amount in the next state. It goes in buyer_location: {"zip": ...}, or {"state": ..., "county": ...} when only the state is known. Without it the car is priced at the store's own location as an estimate (binding: false) and estimate_note says so; the shopper's ZIP firms it up. binding: true means the dealer stands behind the figure. binding: false means it is an estimate the store will confirm, and estimate_note says what is unconfirmed. Every line_items entry says where it came from in source: listing and dealer_profile are the dealer's, and state_table is the tax table for the location in buyer_location. taxes_included: false means the store quotes before government charges: the figure is the price, any discount and the dealer's own fees, with no sales tax, title or registration lines, so it is the price before tax, title and registration and not the full out-the-door price. total is what it costs to drive the car away. A trade-in or a deposit does not reduce it: pass trade_allowance and the tax drops where the buyer's state gives credit for a trade, but what the buyer still owes after their trade is a financing question this tool does not answer. A car CarEdge does not hold for these stores answers not_found. A store CarEdge has no pricing arrangement with answers quote_unavailable: that car can be described but not quoted. quote_id, passed to lead_prepare, records on the lead the figures the shopper actually saw. A car at a dealer in CarEdge's marketplace that is not a CarEdge partner (a rooftop_slug of the form dealer-<id>) gets a CarEdge estimate: the answer has participating: false and is_estimate: true: listing_price is the dealer's own advertised price, not a CarEdge quote; estimate is an out-the-door estimate for the buyer's ZIP (priced_for: "dealer_zip" when none was given), and each of its line_items says in source and detail where its figure came from; dti_grade, dti_score and dti_grade_reason are the dealer's Dealer Transparency Index grade where one exists; caredge_car_url is the car's page on caredge.com and caredge_dealer_url the dealer's, null when it cannot be resolved. A VIN not in the marketplace answers not_found.

stores_list

Claude
Use this when the shopper asks which dealerships are near them, where a dealer is, which store holds a car, which store is closest to them, for a store's address, phone or website, or how transparent a dealer is on price, e.g. near: {"zip": "74133"}. Each store carries dti_grade (A to F) and dti_score (0 to 100) from CarEdge's Dealer Transparency Index, which grades dealers on their fees, their add-ons and how closely the final price matches the listing, from real out-the-door quotes. A dealer that would not give a written out-the-door price is graded F for that reason. Both are null when the store has no grade yet. Do not use it to find cars (that is inventory_search). With near (the shopper's ZIP) it lists dealerships in CarEdge's marketplace within radius_miles, nearest first, each with distance_miles, cars_in_stock, dealer_id and mc_dealer_id. A CarEdge partner has participating: true; any other dealer has participating: false and a dealer-<id> slug: its cars get a CarEdge out-the-door estimate, not the dealer's quote, and no lead. A page has the nearest 50 dealers and all partners in range; past shown_up_to_miles only partners are listed. If more_dealers, send next_offset as near.offset. When these stores span the whole country, near is required, or the answer is the error too_broad. participating is false for a store CarEdge only holds marketplace data for: it can be described, but gives no dealer quote and takes no lead. disclosure_text is the store's own consent wording, null when the store takes no leads or has no consent wording of its own (then lead_prepare returns the wording that applies). When the connector covers one store, sister_stores lists the other stores of the same dealer group, for "do you have another location?". They can be named, but not searched, quoted or sent a lead here.

vehicle_compare

Claude
Use this when the shopper is choosing between two to four cars they have already seen and asks how they differ or which is the better buy, e.g. vins: ["3GNKBKRS6PS118420", "1GNSKRKD2PR339514"]. Do not use it to find cars (that is inventory_search) or for an out-the-door price (that is quote_out_the_door). Each car comes back exactly as vehicle_get gives it, and comparison adds: differs, the fields whose values are not the same on every car (year, trim, engine, drivetrain, color and the like, and store when the cars are at different stores); shared_features and only_features, the features every car lists and, per VIN, the ones no other car lists (a car whose listing has no features is left out of both: its features are unknown, not missing); and lowest_price, lowest_miles and newest, the VIN that wins, or null on a tie or when no car has the figure. price_gap is the spread in dollars. Prices are compared on one basis: price_with_fees (what the dealer's page shows) when every car has it, else the listed price for every car, so a car is never cheaper only because its store's fees are unknown. This is not an out-the-door price; quote_out_the_door gives that for one car. Pass 2 to 4 different VINs. A VIN at any other dealer in CarEdge's marketplace is compared view-only, with participating: false and dealer saying who sells it. A VIN not in the marketplace is listed in missing and left out, as long as two cars remain; otherwise the answer is not_found.

vehicle_get

Claude
Use this when the shopper has picked one car out of inventory_search, or gives a VIN such as vin: "3GNKBKRS6PS118420", and wants to know more about it. Do not use it to search (that is inventory_search) or to price the car out the door (that is quote_out_the_door). photos holds the car's photo links, main photo first, so a chat that cannot show images can still link to them. It is empty when the car has no photos. A VIN at a store this connector covers answers as that store's car. A VIN at any other dealer in CarEdge's marketplace answers view-only. A car at a dealer that does not work with CarEdge comes back view-only: participating: false, no price_with_fees, and a rooftop_slug of the form dealer-<id>. It gets a CarEdge out-the-door estimate, not the dealer's quote, and no lead. dealer says who sells the car: its name, dealer_id, mc_dealer_id, city, state and ZIP, distance_miles from the shopper's ZIP, listed_price (its price), listing_url (the dealer's page for the car, if known) and caredge_url (the car's page on caredge.com). A VIN not in the marketplace answers not_found. When the same VIN is listed at more than one of these stores, the answer is one store's copy and also_at names the others. rooftop_slug picks one store's copy; price and doc fee can differ between stores. participating: false means the dealer gives no quote and takes no lead. source says where the price came from; a marketplace price quotes as an estimate, never as a binding figure. freshness says how old this store's data is, and inventory_unavailable means CarEdge cannot see the lot right now, not that the car is gone. in_transit: true means the car is on its way to the store and not on the lot yet, so it cannot be seen or driven until it arrives. disclosure_text is the store's own consent wording. A lead is accepted only when the shopper has seen this exact wording and agreed to it, and lead_submit gets it back unchanged. It is null when the store takes no leads or has no consent wording of its own; then lead_prepare returns the wording that applies.

Similar Apps

Listed on MCP App Store

Listed on MCP App Store

Add this badge to your site or README. It links to this page.

HTML

<a href="https://mcp-app-store.keenethics-labs.com/app/caredge"><img src="https://mcp-app-store.keenethics-labs.com/badges/listed-on-mcp-app-store.svg" alt="Listed on MCP App Store" width="220" height="44"></a>

Markdown

[![Listed on MCP App Store](https://mcp-app-store.keenethics-labs.com/badges/listed-on-mcp-app-store.svg)](https://mcp-app-store.keenethics-labs.com/app/caredge)