Use Snapy from ChatGPT

What this does

A ChatGPT custom GPT can build a web page for you, but on its own it cannot put that page online. Snapy fixes that. Once you connect it, you can say "build a landing page and host it," and the GPT publishes the page to Snapy and replies with a working link you can open and share. Behind the scenes the GPT calls a single Snapy action, Snapy hosts the file as a static page at a *.snapy.page address, and the live link comes back inside the chat. No copy-paste, no separate hosting account, no deploy step.

This is useful because ChatGPT is good at writing HTML, CSS, and small self-contained pages, but the output normally stays trapped in the conversation. Snapy turns that output into something you can send to a client, a friend, or a phone in one move.

How to set it up

1. In ChatGPT, create a custom GPT, or edit one, and open the Configure tab.
2. Scroll to Actions and click Create new action.
3. Under schema, choose Import from URL and paste: https://snapy.host/openapi.json
4. Save the GPT.

That is it. The action is now available to the GPT. There is no key to paste and no OAuth step, because the Snapy publish endpoint is open and free. If you prefer to wire it by hand instead of importing the schema, the underlying call is a POST to https://api.snapy.host/api/publish with a JSON body containing a content field, and the response carries a url field with the live link.

How to use it

Ask the GPT something like:

Build a simple one page site for my bakery and host it, then give me the link.

The GPT writes the HTML, calls Snapy's publish action, and replies with a live link like bakery.snapy.page. You can open it right away or share it. You can also be specific about the address. Asking for "publish it as fall-menu" tells the GPT to send name: "fall-menu", which produces fall-menu.snapy.page when that name is free.

A few prompts that work well:

Write a thank-you page for my workshop attendees and host it on Snapy.host.
Turn this changelog into a clean HTML page and publish it.
Make a simple pricing comparison table as a web page and give me a link.

Common uses

• One-page sites and landing pages a GPT drafts in seconds.
• Sharing a formatted document (a proposal, a menu, an FAQ) as a real link instead of pasted text.
• Quick previews you send to someone for feedback before building the real thing.
• Resumes, event pages, and simple portfolios.
• Embedding a small interactive widget the GPT writes in plain HTML and JavaScript, since Snapy serves the file exactly as written.

Snapy serves static files. A page that needs a live server, a database, or a backend login will not run on Snapy, so keep GPT output to self-contained HTML, CSS, and client-side JavaScript.

Troubleshooting

• The GPT says it cannot find the action. Reopen the Configure tab and confirm the action imported from https://snapy.host/openapi.json without errors. Re-import if the schema list is empty.
• No link came back. The publish call needs a JSON body with a content field. If the GPT sent an empty or malformed body, ask it to "publish the full HTML as the content field" and try again.
• The name was rejected or you got a random address. That name was already taken. Pick another name, or just omit the name and Snapy assigns a random one.
• You wanted a PDF or image hosted. ChatGPT actions are best for text and HTML. To host a binary file, base64 encode it and send it as content_base64 over the API. The Python guide shows that pattern.

Good to know

• The link is free and needs no account.
• You get a private analytics link too, so you can see views later.
• You can add a password or an expiry to the link afterward on Snapy.

Want the technical details? See the developer docs, browse the other integrations, or read how to share an AI-generated page. If you also use Claude, the Claude guide covers the MCP connector.

Connect it once and every page your GPT builds becomes a link you can actually open.