TL;DR. Install Claude Code, add Hatchable as an MCP server with one command, then ask Claude to build and deploy your app in plain English. You'll get a live URL in a few minutes. The whole flow is free — bring your own Claude subscription.
What this gets you
By the end of this guide, you'll have a working app deployed at https://your-project.hatchable.site, with:
- Static files served from a CDN-backed URL
- A Postgres database attached to your project (no setup)
- File storage for uploads / generated assets
- Auth you can turn on per-route
- Zero config files, zero build pipeline to learn
Claude Code does the coding. Hatchable does the hosting. You describe what you want.
What you need
- Claude Code installed.
npm install -g @anthropic-ai/claude-code, thenclaude loginto connect your Anthropic account. Free-tier Claude works; a paid plan just gives you more messages. - Node.js 20+. Required by Claude Code itself.
- A Hatchable token. Free, no signup form — grab one at hatchable.com by clicking Copy on the Claude Code snippet. The command includes a fresh token.
Step-by-step: from prompt to live URL
-
Add Hatchable as an MCP server
Paste the command from hatchable.com — it looks like:
claude mcp add --transport http hatchable https://hatchable.com/mcp \ --header "Authorization: Bearer YOUR_TOKEN"This registers Hatchable with Claude Code. Verify it worked with
claude mcp list— you should seehatchablein the output. (Stuck? See our full MCP setup guide.) -
Start a Claude Code session in an empty directory
Make a new folder, open it in your terminal, and run:
mkdir my-first-app && cd my-first-app claudeThe working directory matters — Claude Code treats it as the project root. Start fresh so Claude isn't distracted by unrelated files.
-
Describe the app
Prompt Claude in plain English. The more specific you are about what the app does, the better the first draft. Example:
"Build a habit tracker on Hatchable. Let me add habits with a name and a color. Show them on a grid with a checkmark for each day in the current month. Save everything to the database. Make it look clean — warm, cream background, serif headings."
Claude will respond with a plan (what files it'll create, what tables it'll add to the database, what the UI will look like) and ask for your approval to proceed. Approve it.
-
Let Claude work
Claude will call Hatchable's MCP tools in sequence — typically:
create_project— registers a new project on Hatchable and picks a subdomainwrite_files— writes your HTML/CSS/JS and any API routesrun_migration— creates the database tablesdeploy— pushes the current project state to live
Each tool call may prompt you for permission the first time (Claude Code's default). Accept them, or launch with
claude --allowedTools "mcp__hatchable__*"to pre-approve all Hatchable calls. -
Open the live URL
When Claude's done, it'll print the live URL — something like
https://my-first-app.hatchable.site. Open it; you're looking at the app Claude just built and deployed. Takes 1-3 minutes end to end for a simple app, 5-10 for something more involved.
What if something breaks?
Claude can't find the Hatchable tools
Exit Claude (Ctrl+D), run claude mcp list — Hatchable should be listed as connected. If not, re-run the claude mcp add command with a fresh token from the homepage. Hatchable tokens expire after 30 days of inactivity.
Deploy fails with "project over quota"
Unclaimed anonymous accounts get 1 free project. If you've already deployed one, either claim the existing account (click the claim prompt the next time you visit your live app) or generate a fresh token. Claimed accounts can have unlimited personal projects.
The app is live but looks broken
Open the browser DevTools console and tell Claude what error you see. Claude will read the error, find the bug, and ship a fix. This iteration loop is the whole point — you're not supposed to read the code; you're supposed to describe what's wrong with the running app.
I want to edit code myself
Everything Claude writes lives in your project directory on disk. Open it in any editor, change what you want, then ask Claude to deploy the changes (or run claude mcp call hatchable deploy directly if you prefer).
What comes next
- Share it privately. By default your app is personal — accessible only to you. Add collaborators from the Hatchable console to let specific people in.
- Make it public. Upgrade to Pro ($12/mo) and the app can be published to the open web with a custom domain.
- Build another. Each Hatchable project is independent. Start a new Claude Code session in a new directory and repeat.
Ship your first Claude Code app today.
Free hosting with database, auth, and storage. No credit card, no server setup.
Get started free →Frequently asked questions
How long does a typical deploy take?
1-3 minutes end-to-end for a simple app from scratch: Claude writes the code, calls Hatchable's deploy tool, files go live. Subsequent redeploys of the same project are faster — usually 30-60 seconds.
Can I deploy without using Claude Code?
Yes. Hatchable's MCP server works with any MCP-compatible AI tool — Cursor, Codex, Antigravity, OpenClaw. The deploy experience is nearly identical; only the client configuration differs. See our Cursor guide or Codex guide.
Does Hatchable charge for Claude usage?
No. Hatchable never touches Anthropic's billing. You use your own Claude account (free or paid) for the AI, and Hatchable is free for the hosting. The two are completely separate.
What if my app needs a feature Hatchable doesn't support?
Hatchable covers the common web-app stack: static files, HTTP APIs, Postgres, file storage, email, and auth. Things it doesn't do yet include real-time (WebSockets), heavy media transcoding, and native mobile apps. For those, Claude Code can still write the code — you'd just host the result elsewhere.
Can I use a custom domain?
On the free plan, apps are served at {slug}.hatchable.site and are accessible only to you and invited collaborators. Upgrade to Pro ($12/mo) to publish to the open web at your own domain.