the cloud your AI builds on.
Ship real apps, just by asking.
Hosting, databases, files, email and sandboxes behind one API your agent already knows how to use. You decide what it can touch, and what it can spend.
Agent session
mcp · opencloudportPut my bakery’s order form online, with a database for orders and email receipts.
create_appname: "Bakery orders"Runningapp_7Qx… createdadd_servicetype: "compute"RunningReadyadd_servicetype: "database"Runningenv.DB readyadd_servicetype: "email"Runningenv.EMAIL readyquery_databasecreate table orders (…)Running1 tabledeploymain: "index.js"RunningLive
Your order form is live. Each order is saved to env.DB, and a receipt goes out from the app’s own address.
opencloudport is a cloud for AI agents and the people who use them. Ask for an app in plain words and your agent puts it online, with hosting, a database, files, email and sandboxes behind one connection. Works with Claude Code, Cursor, your own agent and your own code.
claude mcp add --transport http opencloudport https://opencloudport.com/mcp$ claude
◆ What should we build?
An order form for my bakery, with email receipts
◆ Plug in
● Compute
● Database
○ File storage
⚙ create_app · add_service × 3 · query_database · deploy
✓ Live. Each order is saved to env.DB and a receipt goes out.
Works with the agent you already use.
Claude Code, Cursor, or an agent you built. One line connects it, and it gets a tool for every job, each described in plain words.
Calm for you. Fast for agents.
Give every agent its own key. Keep it to one app, make it read-only, and set what it may spend. Everything it changes is written down in plain words.
› deploy
402 key_spend_limit_reached
This key reached its $5.00 spending limit. Raise or remove it on the API keys page.
- A key per agent
- Spending limits
- An activity log
- Asks before deleting
Anybody can build.
Use the dashboard, hand it to your agent, or write the code yourself. It’s the same cloud, the same keys, the same bill.
Everything on one page.
Make apps, plug in what they need, and look inside them, without writing code.
- A map of every app and how its parts connect
- A console to look inside your databases
- A file browser, and a terminal for sandboxes
- Usage in plain words, like “2 h 14 min of Medium sandboxes”
Everything an app needs.
Apps start empty. Plug in what each one needs, as many of each as it needs, and its code reaches every part by name.
Compute
Runs your code and answers at your app’s own address. Roll back in one click.
Databases
SQL tables your code reaches as env.DB, with a console to look inside.
File storage
Uploads and images, sent straight from a visitor’s browser with signed links.
Receipts and sign-ups, sent from your app’s own address.
Sandboxes
Private Linux machines where agents install, build and preview code. Free while asleep.
Keys with limits
Give each agent its own key: one app, read-only if you like, with a spending limit.
Pay for what runs.
One prepaid balance.
Usage is charged at the providers’ own prices, nothing added, and follows them when they change. Our fee is 5.5% (at least $0.80) when you add funds. No plans, no seats.
- Prices follow the providers, the same day
- Can’t spend more than your balance
- Apps pause after 24 hours at $0, and nothing is deleted
Build your apps
by asking, in plain words.
Give your agent a cloud.
Create your account, add a balance, and your first app can be live today.
Prepaid. You can’t spend more than you put in.
Yours. Your data stays put, even when a balance runs out.
In minutes. From an empty app to live, by asking.
opencloudport