Docs  /  Getting started
01 — Getting started

From idea to live URL in five minutes.

No terminal, no build tools, no infrastructure to manage. Connect the AI you already use, tell it what you want in plain English, and it builds it. When it's ready you'll have a real web address you can share.

1

Make an account

Head to hatchable.com and sign up with your email. No credit card. The free tier covers everything you need to build, test, and share with people you invite, plus your first app on the open web. You only pay when you want to publish more than one thing, or put it on your own domain.

If you've already got an account, you're done with this step. Open your console →

2

Connect your AI to Hatchable

Hatchable doesn't have its own chat window. You build by pointing an AI tool you already use (Claude, Claude Code, ChatGPT, Codex, Antigravity, or Cursor) at Hatchable, then talking to it the way you always do. Don't have one yet? Claude is the quickest to set up, and its free tier is plenty to start.

Hatchable speaks MCP, a small open standard that lets AI tools call into other tools safely. Adding Hatchable takes one paste. After that, your AI can create projects, write code, deploy them, and check that they work, all in the same conversation.

Claude & Claude Code · one-click setupAdd Hatchable to Claude → (URL + name pre-filled; click Add, then Connect on the row to start OAuth. One setup covers both surfaces.) ChatGPT · listed in the app directoryOpen Hatchable in ChatGPT → (click Connect on the listing; no Developer Mode, works on every plan) Codex · install the pluginOpen Plugins in the Codex app (CLI: run /plugins), install Hatchable, approve the sign-in. Antigravity · ask the agent in chatAdd the https://hatchable.com/mcp MCP connector with OAuth. Cursor · one-click setup (Cursor 1.0+)Add Hatchable to Cursor → (OAuth handled on first connect, no token) Cursor (older versions) · ~/.cursor/mcp.json{ "mcpServers": { "hatchable": { "url": "https://hatchable.com/mcp", "headers": { "Authorization": "Bearer $HATCHABLE_TOKEN" } }}}

Every client here signs you in through a browser tab. Once the connector is added, click Connect (in Codex it's the plugin's sign-in prompt; in Cursor and Antigravity it's an OAuth prompt on first use), approve on hatchable.com, and you're done. Claude Code shares its connector list with Claude on the web, so one setup covers both. Only older Cursor versions (before 1.0) skip OAuth: they need a static token from your API keys page in place of $HATCHABLE_TOKEN.

3

Tell the AI what to build

With Hatchable connected, you just talk to your AI like normal. Describe the thing, who it's for, what matters most. Don't worry about being technical. The more specific you are about the use case, the better the first version comes out.

You
Build me a chore tracker for my kids on hatchable. Each kid has their own page where they can check off their daily chores. I want to see a weekly summary on my page showing who finished what.
→ I'll set up a small database with kids and chores, build a kid view and a parent view, and add the weekly summary. Give me about a minute.
The two-word signal: "on hatchable"

Tacking "on hatchable" onto the end of any prompt tells your AI to reach for the Hatchable tools rather than writing local files. It's the difference between "I'd like to talk about an app" and "please build this app right now." Use it once at the start of a session and your AI usually sticks with the platform from there.

The AI will ask a question or two if something's ambiguous ("how many kids?", "should chores reset weekly or daily?"). Answer in plain English. It plans the whole thing, writes the code, sets up the database, and tests its own work before showing you anything.

4

Watch it deploy

You'll see a stream of "doing this, doing that" go by: making the database, writing pages, testing. When it's done you get a live URL like your-name.hatchable.site/chores. That's a real, working website you can open on your phone right now.

What you get out of the box

  • A real web address with HTTPS, private to you until you choose to publish
  • A private database that holds your data
  • Login, if your app needs it (your kids signing in, etc.)
  • Email sending, if your app needs to send any
  • File uploads, scheduling, and more, only the things your app actually uses

All of it included. Nothing to configure. Nothing to bolt on. Hatchable handles every piece of plumbing so the AI can focus on building what you asked for.

5

Share it (or keep it to yourself)

By default your app is private: only you can see it. To bring others in, invite them by email or hand them a single-use invite link, then pick their access. Member lets someone use the app; Admin lets them edit it alongside you. They sign in and they're in. No seat limits. Invite your family, your team, your whole book club.

Want it out on the open web, no login required? That's a one-click flip in the project settings, and your first public app is free. See Sharing & selling for how the plans work.

6

Change anything by talking

Forgot a feature? Spot a bug? Want a different color scheme? Go back to your AI and say so. The AI rewrites the relevant pieces and redeploys. Every change is tracked, so if a new version breaks something you can roll back with one click.

You, two days later
Add a "missed chore" badge that shows up next to a kid's name when they've skipped something for two days in a row.
→ Done. Took 22 seconds. Open the parent view to see the new badge. Maya's got one already.

You're never stuck with the first version. The whole point is that you keep talking to it, and the app keeps getting closer to what you actually wanted.

Where to look in your console: the Code tab has a Files view and a History view, where each change is a labeled entry: the AI's request as the title, its plain-language summary as the body. Once you publish your site to the web, new changes land as drafts at your-app-draft.hatchable.site first, where only you and your collaborators can see them. You click Promote to live when each one's ready. Your live site stays untouched until you do.

That's the whole thing.

If you've got an idea right now, even a fuzzy one, you can have a working version of it before lunch.

Start building