What is Placeground?
The short version: you describe it, Placeground builds and hosts it.
Placeground is a home in the cloud for the things your AI assistant makes for you. Ask for a website, an app, a place to keep data, or sign-ins for your users — and it becomes real and live on the internet, without you touching any of the plumbing.
The problem it solves
Normally, turning an idea into a working app means a wall of technical steps: renting servers, wiring up a database, setting up accounts, and deploying — each one a place to get stuck. Most of that work has nothing to do with your actual idea.
Placeground takes care of all of it. Your assistant talks to Placeground directly, so the moment something is built it's already running in the cloud — no servers to manage, no setup, no deploy steps.
Focus on creating
That's the whole philosophy: you stay in the part that matters — imagining what you want and describing it — while Placeground handles everything underneath. Your creativity never has to pause so you can learn infrastructure. You create; it takes care of the rest.
01Connect your assistant
A one-time setup. After this, you just talk to your assistant.
1. Add Placeground
Placeground connects to your assistant through MCP — a small standard that lets any AI client talk to the same service. You set it up once per assistant. Pick yours below; they all point at the same address: https://api.placeground.cc/mcp.
Claude Code
Run the command below, then restart Claude, type /mcp, pick placeground, and finish the sign-in that opens in your browser. The tools appear once it shows connected.
OpenCode
Add this to opencode.json (or ~/.config/opencode/opencode.json), then complete the browser sign-in OpenCode prompts you with.
{
"$schema": "https://opencode.ai/config.json",
"mcp": {
"placeground": {
"type": "remote",
"url": "https://api.placeground.cc/mcp",
"enabled": true
}
}
}
Codex
Open Settings → MCP servers → Add new server, choose Streamable HTTP, paste the URL below, then finish the browser sign-in Codex opens.
Cursor
Add this to ~/.cursor/mcp.json (or .cursor/mcp.json in your project), restart Cursor, then authorize placeground in your browser from Settings → MCP.
{
"mcpServers": {
"placeground": {
"url": "https://api.placeground.cc/mcp"
}
}
}
Claude Cowork
Open Settings → Connectors → Add custom connector, paste the URL below, and finish the browser sign-in. Cowork runs from Anthropic's cloud, so only remote endpoints like this one work — there's nothing to install.
Any other MCP client
Any MCP-compatible assistant connects over the same Streamable HTTP endpoint. Point your client at the URL below and complete the browser sign-in it opens — no API key to paste.
2. Sign in once
A browser tab opens so you can sign in — just like logging into any website. You only do this the first time. After that, your assistant stays connected for you and you never log in again.
3. Ask for what you want
That's it. Describe something in your own words and your assistant builds it. Everything it makes is saved under your account and stays yours — your apps and data don't disappear.
02How it works
You talk; your assistant uses Placeground behind the scenes.
Placeground gives your assistant a set of building blocks — a place for accounts, a place for your data, a place for your files, and a real web address where your app goes live. When you ask for something, your assistant picks the right pieces and wires them together for you.
Everything you make is organized into apps. Each app is its own private space with its own data and files — nothing leaks between them. You can have as many as you like.
Two things you can build
- A website — a page anyone can open and share (a landing page, a portfolio, a "hello world"). It goes online instantly. → Build a website
- A full app — something with logins and saved data (a to-do list, a shop, a booking tool). → Build a full app
You don't have to know which is which — just describe what you want and your assistant chooses. The sections below show what to say for each.
03Build a website
A page someone can just open. Live the moment it's built.
Ask for a page or a site and your assistant writes it and puts it online right away. You get a clean link to share — there's no build step and no "deploy" button to press.
- Builds the page
- Puts it online for you
- Hands you the link
Changing it later
Just ask. Describe the change in plain words and your assistant updates the page in place — the same link keeps working.
- Updates the page
- Re-publishes it instantly to the same address
04Build a full app
For anything with logins and saved data.
When you want people to sign in, save things, or come back to their own data, ask for an app. Your assistant sets up the accounts, decides where the data lives, builds the pages, and puts it all online — in one go.
- Creates a new app
- Adds sign-up & log-in
- Sets up where tasks are saved
- Builds the pages and puts them online
05Your data
Save and find information — without ever setting up a database.
Your app remembers things in collections — think of them like simple tables (products, tasks, orders). You don't create or configure anything; your assistant organizes it for you the first time you mention it.
- Sets up a "products" collection (once)
- Saves each product
Finding and changing things
Ask for exactly what you need — filtered, sorted, counted, updated.
- Reads your orders
- Filters by date and amount, sorts by newest
- Shows you the results
06Files & images
Upload pictures and documents. Share some, keep others private.
Add images and files to your app. Public ones get their own shareable link right away; private ones stay hidden and only reachable through a short-lived link you ask for.
- Uploads the logo as public and uses it on the site
- Uploads the receipts as private
07Accounts & sign-in
Let the people who use your app sign up and log in.
This is different from your Placeground account. It's the sign-in for the visitors of the app you build — so each person sees only their own data. Your assistant sets it up; you just ask.
- Adds sign-up & log-in for your visitors
- Ties each note to the person who created it
- Makes sure nobody sees anyone else's notes
08Going live
It's already online — there's no deploy step, ever.
The moment your assistant builds something, it's live. Every site comes with two links:
- A share link — clean and stable, the one you give to other people (e.g. https://taskbird.placeground.site).
- A live-preview link — the same site, but it refreshes by itself every time you make a change, so you watch your app update as you ask for tweaks.
Every link is served securely over HTTPS — the SSL certificate is set up and renewed for you automatically, so the secure padlock is on from the very first second. There's nothing to buy and nothing to configure.
- Makes the change
- Re-publishes to the same links instantly
- Your open preview refreshes on its own
09Custom domains
Put your app on a domain you own instead of the placeground.site link.
Already have a domain like yourname.com? Ask to connect it. Your assistant hooks it up and tells you the two small records to add at your domain provider — one that points the domain at your app, and one that turns on the secure padlock (HTTPS).
- Connects the domain to your app
- Gives you the records to add at your domain provider
- Waits for them, then turns on HTTPS automatically
The domain shows pending until those records take effect (usually a few minutes), then it goes live on its own. You can also ask which apps your domains point to, or disconnect one.
10For developers
What's happening under the hood — and how to build on it.
Everything above is built on a small, predictable architecture. You can read, edit and re-run all of it; the local project your assistant writes is the source of truth.
Architecture
A static frontend (a Next.js static export) talks to server-side functions over HTTP. The browser never touches the database or storage directly — collections and storage are server-only. All business logic lives in functions, which your assistant writes for you.
Functions
One file per endpoint, named <resource>-<action> (e.g. tasks-create). Each file defines a single handler:
// functions/tasks-create.js async function handler(request, pg, ctx) { // AUTH — require a signed-in end-user if (!ctx.user) return new Response("Unauthorized", { status: 401 }); // VALIDATE — read and check the input const { title } = await request.json(); if (!title) return { error: "title is required" }; // WORK — save it, scoped to this user return pg.collection("tasks").create({ title, userId: ctx.user.euid }); }
pg is a data client bound to the app (pg.collection(name).create/get/list/query/count/update/mutate/delete, pg.storage) and can never touch another app. ctx is { appId, user, url, method }; keeping each user's data private (filtering by ctx.user.euid) is up to your function.
Data model
Collections are typed tables. Declare fields before adding records — types are string, number, boolean, datetime, json and relation. Mark fields you filter or sort on with index: true. Adding fields later is fine; changing a field's type is not.
Keys
- Publishable key (pk_) — browser-safe, but powerless on its own: it can only sign end-users in/out and invoke functions. Ship it in the frontend.
- Secret key (sk_) — full admin, bypasses everything. Keep it server-side and in setup scripts only; never put it in the browser.
Deploying & editing
There's one way to deploy: the pg CLI (pg deploy ./out --app <appId> --id <deploymentId>). It hashes every file, uploads only what changed, and prints the live URL. Reuse the same deployment id across re-deploys so the links stay stable. To tweak an already-deployed site without local files, your assistant can read a file from the live site and write it back in place.
11FAQ
Do I have to deploy anything?
No — never. The moment your assistant builds your app, it's already online with its own link. There's no build step and no "deploy" button to press.
Do I need to know how to code?
No. You describe what you want in plain words and your assistant does the rest. If you do know how to code, you can read and build on top of everything it makes.
Does it only work with Claude?
No. It works with any AI assistant that can connect to an MCP server. Just give it the address and sign in once.
Do I have to log in every time?
No. You sign in once, the first time you connect. From then on it's automatic — you won't have to log in again.
Is my data private?
Yes. Each app keeps its data separate, and only you — and the people who sign into your app — can reach it. Private files don't even show up to anyone else.
Can I use my own domain?
Yes. Ask your assistant to connect a domain you own; it gives you a couple of records to add at your domain provider, then turns on HTTPS automatically. See Custom domains.