Back to tutorials
Tutorials

How to design a Tesla wrap from Claude or ChatGPT

By FrunkLabSeptember 10, 20266 min read

You can now ask your AI assistant for a Tesla wrap. FrunkLab runs a server that speaks the Model Context Protocol, the standard Claude, ChatGPT, Cursor and Claude Code use to talk to outside tools. Connect it once, and "design a matte black wrap with copper accents for my Model Y" becomes a finished Paint Shop wrap in your FrunkLab account, previewed on the 3D car, with the download one click away.

This tutorial walks through the connection on each assistant, what the approval screen is asking, what the assistant can and cannot do with your account, and how to cut it off again. It takes about two minutes.

What you need

  • A FrunkLab account with a confirmed email address. Sign up at frunklab.com/signup if you do not have one; the connection is refused for unconfirmed accounts.
  • An assistant that supports custom MCP connectors. That is Claude on every plan (the free plan allows one custom connector), ChatGPT on Plus, Pro, Business or Enterprise with developer mode switched on, Cursor, or Claude Code.
  • The server address: https://www.frunklab.com/api/mcp. It is the same for everyone.

Generations made from an assistant spend the same monthly allowance as the website. A free account gets 3 wrap generations and 2 sound-effect chime generations a month, plus 5 free melody or voice chimes a day; the pricing page lists the paid plans. Nothing is charged for a failed generation.

Step 1: Add FrunkLab as a custom connector

In Claude. Open Settings, then Connectors, then Add custom connector. Paste the server address and confirm. Claude fetches the connection details on its own.

In ChatGPT. Open Settings, then Connectors, then Advanced settings, and turn on Developer mode. Back on the Connectors screen a Create button appears. Paste the server address, pick OAuth as the sign-in method, and save.

In Claude Code. In a terminal, run:

claude mcp add --transport http frunklab https://www.frunklab.com/api/mcp

Then type /mcp inside Claude Code to start the sign-in.

In Cursor or another client. Add the server address to the client's MCP settings. Clients that run the OAuth flow themselves work like Claude; the rest accept an API key instead (see the last section).

Step 2: Approve the connection with your FrunkLab account

The assistant opens a FrunkLab page in your browser. If you are not signed in, you sign in first, with the same email and password or Google account you always use. The approval page then shows three things:

  1. Which app is asking. The name the assistant registered with, such as Claude.
  2. Where you will be sent afterwards. For Claude and ChatGPT that is their own site. For Claude Code and Cursor it is a program on your computer, and the page says so in an amber box; only approve that when you started the connection yourself a moment ago.
  3. What the assistant will be able to do. Generate wraps with your allowance, generate lock chimes, see your designs, and read your plan and remaining generations. It never sees your password, email or payment details.

Tap Approve. The browser returns to the assistant and the connection is live. The whole approval is a standard OAuth flow: the assistant holds a short-lived key that only works at FrunkLab's MCP server, and it can be revoked at any time.

Step 3: Ask for a wrap

Talk to the assistant the way you would talk to a friend who designs wraps:

Design a wrap for my Cybertruck: desert camo in sand and rust with orange accents.

Behind the scenes the assistant looks up the exact vehicle, calls the generator once, and gets back the image, a link to the design page and how many generations you have left. Good assistants show the image inline; the rest show the link. You also get an email with the link, so the wrap is never lost in a chat.

A generation takes ten to forty seconds. If the assistant reports the wrap is "pending", it is still rendering; it appears in My Designs within a minute and the assistant should not ask again.

Prompts that work follow the same rules as the AI wrap generator on the website: colours, textures, gradients and mood over long stories, and no logos or legible text, which the generator will not paint. Our prompt-writing tutorial has a dozen examples.

Step 4: Open the design and send it to the car

The design page shows the wrap on a rotating 3D model of your Tesla. From there you can download the Paint Shop PNG, edit the design in the studio or the Garage, attach a lock chime, and submit it to the gallery.

To put it on the car, upload the PNG from the Tesla app (Creations, then Wrap, then Upload) or copy it to a USB drive. The Tesla app upload tutorial covers both routes and the ten-wrap limit on the car.

Step 5: Ask for a matching lock chime

The same connection generates lock chimes:

Make me a lock chime that goes with it: a short desert-wind sting with a two-note metallic chime.

The assistant returns a download link for a Tesla-ready LockChime.wav, valid for an hour, and the chime is saved to your My Chimes library either way. Put the file at the root of a USB drive and switch it on in the car under Controls, then Locks. Spoken chimes work too: put the words in quotes, as in "a calm voice saying 'doors locked'".

What the assistant can do, and what it cannot

It can: list every Tesla model FrunkLab supports, tell you how many generations you have left, generate wraps and chimes, list your designs, search the 400+ lock sounds, and browse popular gallery wraps.

It cannot: delete anything, change your account, see your email or billing, spend money, or touch anyone else's designs. When your monthly allowance runs out it tells you the reset date and where to read about plans; it never starts a purchase.

Disconnecting, and using an API key instead

Open your account page and find Connected apps and API keys. Every connected assistant is listed with a Disconnect button. Disconnecting takes effect on the assistant's next request.

The same card creates API keys for clients that cannot run the sign-in flow, such as Claude Code with a fixed header, some Cursor setups, or the OpenAI Responses API. A key is shown once when created, so copy it straight into the client. In Claude Code:

claude mcp add --transport http frunklab https://www.frunklab.com/api/mcp --header "Authorization: Bearer fl_ak_YOURKEY"

Treat a key like a password: it can generate with your allowance until you revoke it from the same card.

Ready to try it? Open your assistant's connector settings, paste https://www.frunklab.com/api/mcp, and ask for a wrap. The AI assistants page has the same steps in short form for reference.

Frequently asked questions

Does the assistant see my wrap image? It gets the image address and the design link. Assistants that render images show it inline; others show the link. You also get an email with the link.

Can I use FrunkLab from an assistant without an account? No. Every generation belongs to an account so it lands in your designs and counts against your allowance, and the approval screen requires a confirmed email.

Does a generation from the assistant cost the same as one on the website? Yes. It is the same allowance and the same tokens. A failed generation is never charged.

Can I edit a wrap the assistant made? Yes. Open the design page and choose Edit. It opens in the studio or the Garage like any other design.

#ai#mcp#claude#chatgpt#how-to#custom-wrap#lock-chime

Put a custom wrap on your Model Y

Community designs, free to grab — or make your own in the browser.

Kiwaieeee - Model Y Juniper Performance — custom Tesla wrap designMoney Maker — custom Tesla wrap designBenjamins' Ride — custom Tesla wrap design