MCP App Store
by KeenEthics
CarClever – Buying Assistant icon

CarClever – Buying Assistant MCP Server for Claude and ChatGPT

By Broekman Consulting Pty Ltd

Overview

Find your next car with a buying assistant that helps you search, check and compare real U.S. vehicle listings. CarClever searches a daily refreshed Edmunds vehicle catalogue to build a shortlist around your needs and budget. Search by location, make, model, year, trim, price, mileage, colour, drivetrain (where the VIN decode confirms it) and electrification, or describe practical needs such as a family SUV, commuter car or electric truck. Prioritize the lowest price, newest model year or lowest mileage. Ask for best value to rank cars by their price against similar local listings (a ZIP code is needed). Ask for more of the same search, up to 16 cars in total. Already found a car? Check its VIN for vehicle identity, available listing details and NHTSA recall information. See model-year safety ratings and equipment information where available, with clear distinctions between confirmed details and items to verify with the dealer. Before you contact a dealer, you also get what is still unknown and the questions worth asking. Compare two to four vehicles side by side on price, mileage, age, trim, powertrain, location and available NHTSA information. Newly observed listings and recorded asking-price drops help you spot options worth another look. Each listed vehicle includes an Edmunds link. Open the listing and contact the dealer to confirm availability, the out-the-door price, equipment and your next step. If an exact listing is unavailable, CarClever can provide a search for similar vehicles. Vehicle listings come from Edmunds through its Impact affiliate catalogue; vehicle decoding and recall information come from NHTSA. Vehicle links open on Edmunds through affiliate links. CarClever may earn a commission from qualifying enquiries. Results are not ranked by commission.

How to connect

  1. Claude: Open the CarClever – Buying Assistant listing in the Claude connectors directory and choose Connect. Claude listing CarClever – Buying Assistant exposes 5 tools here.
  2. ChatGPT: Open the CarClever – Buying Assistant listing in the ChatGPT app directory and choose Connect. ChatGPT listing CarClever – Buying Assistant exposes 5 tools here.

Tools

check_vehicle

ChatGPT
A VIN check for one known vehicle: identity, recall and listing details, for a VIN, or for a make, model and year. For a used car with a VIN, it includes NHTSA's recall lookup link. Use it for one known vehicle, including a follow-up to a search. To compare two or more vehicles, use Compare Vehicles. For a broad search, use Find Matching Vehicle.

compare_vehicles

ChatGPT
Compares 2 to 4 vehicles side by side on price, mileage, year, trim, fuel, drive, engine, NHTSA safety rating, recall campaign count and distance (with a ZIP). Takes VINs. Use it whenever the user asks to compare vehicles, including "compare 1 and 2" after a search: pass the VINs of those results from the earlier search.

find_matching_vehicle

ChatGPT
Finds US listings (updated daily) for new or used vehicles that match the buyer's stated requirements, ordered by the requested objective (cheapest, newest, lowest mileage, best within budget). Use it for explicit criteria or practical needs (a family SUV, a commuter car, a tow vehicle), or an exact listing by VIN. Not for car education, maintenance, financing or leasing. Rules: 1. Pass only what the buyer stated or clearly implied. Set condition to new or used when the buyer says so; leave it out when they say neither. Check the reply's Searched line against the request and search again if they differ. 2. Make, model, trim, year, price, mileage, location, condition, colour, body style and doors are matched against listing data. Trim preference ranks only; seats are not a filter. Transmission and cylinders are not filters; results show them when known. 3. Keep any models the buyer named. For a broad search, choose real models that fit the size, class or need, and put the need in vehicleNeeds. Say the models are your picks for the size, and point the buyer to the listing for dimensions. If results are thin, add more suitable models. Do not copy examples from this description. 4. For hybrid, plug-in or electric requests, pass the accepted types and whether they are required or preferred. Pass base models that offer that version; the tool also searches their electrified versions. NHTSA VIN data decides: for required searches, confirmed cars come first, cars NHTSA cannot confirm are kept and marked, and cars it shows as another type are left out; preferred searches only rank. A model family offering a hybrid, or a fuel label, does not make a listing one. 5. Use trimRequired for a trim the buyer requires, trimPreference for an optional one. Never infer a trim from a seat count or another need; leave it unset when unsure. 6. Drivetrain, electrification and seating come from NHTSA data; anything it cannot confirm is marked. Each result links to Edmunds; changed criteria are stated. 7. For a city-only request, use a valid representative ZIP for that city. Reuse an earlier location only when continuing the same search; never assume an unrelated one. Do not rerun with a larger radius unless the buyer asks or agrees; if local results are thin, offer a wider search. Report the actual scope and any location fallback. 8. Report only what the results show; do not present inferred models or trims as confirmed. Missing data stays unknown and never proves a car meets or fails a request. For anything not shown (used-car history, certification, reliability, safety, running cost, towing suitability, physical condition), say where to look: the Edmunds listing and the dealer. Do not describe what the tool lacks. Do not mention history for new cars. 9. A VIN refers to one listing. If it is not listed, say so and link to similar vehicles without substituting another car. If lookup is unavailable, say so; that is not proof the car is absent or sold.

show_more_vehicles

ChatGPT
Shows more cars from a search you already ran. Use the reference printed under that search's results and the total number of cars to show, up to 16; the cars already shown keep their numbers and new ones are added after them. Only for a search that printed a reference: never make one up. If the list has expired or the reference is unknown, run a new search.

check_vehicle

Claude
A VIN check for one known vehicle: identity, recall and listing details, for a VIN, or for a make, model and year. For a used car with a VIN, it includes NHTSA's recall lookup link. Use it for one known vehicle, including a follow-up to a search. To compare two or more vehicles, use Compare Vehicles. For a broad search, use Find Matching Vehicle.

compare_vehicles

Claude
Compares 2 to 4 vehicles side by side on price, mileage, year, trim, fuel, drive, engine, NHTSA safety rating, recall campaign count and distance (with a ZIP). Takes VINs. Use it whenever the user asks to compare vehicles, including "compare 1 and 2" after a search: pass the VINs of those results from the earlier search.

find_matching_vehicle

Claude
Finds US listings (updated daily) for new or used vehicles that match the buyer's stated requirements, ordered by the requested objective (cheapest, newest, lowest mileage, best within budget). Use it for explicit criteria or practical needs (a family SUV, a commuter car, a tow vehicle), or an exact listing by VIN. Not for car education, maintenance, financing or leasing. Rules: 1. Pass only what the buyer stated or clearly implied. Set condition to new or used when the buyer says so; leave it out when they say neither. Check the reply's Searched line against the request and search again if they differ. 2. Make, model, trim, year, price, mileage, location, condition, colour, body style and doors are matched against listing data. Trim preference ranks only; seats are not a filter. Transmission and cylinders are not filters; results show them when known. 3. Keep any models the buyer named. For a broad search, choose real models that fit the size, class or need, and put the need in vehicleNeeds. Say the models are your picks for the size, and point the buyer to the listing for dimensions. If results are thin, add more suitable models. Do not copy examples from this description. 4. For hybrid, plug-in or electric requests, pass the accepted types and whether they are required or preferred. Pass base models that offer that version; the tool also searches their electrified versions. NHTSA VIN data decides: for required searches, confirmed cars come first, cars NHTSA cannot confirm are kept and marked, and cars it shows as another type are left out; preferred searches only rank. A model family offering a hybrid, or a fuel label, does not make a listing one. 5. Use trimRequired for a trim the buyer requires, trimPreference for an optional one. Never infer a trim from a seat count or another need; leave it unset when unsure. 6. Drivetrain, electrification and seating come from NHTSA data; anything it cannot confirm is marked. Each result links to Edmunds; changed criteria are stated. 7. For a city-only request, use a valid representative ZIP for that city. Reuse an earlier location only when continuing the same search; never assume an unrelated one. Do not rerun with a larger radius unless the buyer asks or agrees; if local results are thin, offer a wider search. Report the actual scope and any location fallback. 8. Report only what the results show; do not present inferred models or trims as confirmed. Missing data stays unknown and never proves a car meets or fails a request. For anything not shown (used-car history, certification, reliability, safety, running cost, towing suitability, physical condition), say where to look: the Edmunds listing and the dealer. Do not describe what the tool lacks. Do not mention history for new cars. 9. A VIN refers to one listing. If it is not listed, say so and link to similar vehicles without substituting another car. If lookup is unavailable, say so; that is not proof the car is absent or sold.

show_more_vehicles

Claude

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/carclever-buying-assistant"><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/carclever-buying-assistant)