Skip to content

Beta: these docs are still being updated. The next full update is on 18 November at 8 PM UTC. Docs in beta. Full update 18 Nov, 8 PM UTC.

Hoody.com

You don’t need to learn Hoody before you use it. Your AI can learn it for you. Type one line into the AI you already use, and it signs you in, gets your computer ready and does the task you give it.

Paste this into your AI’s chat:

@https://hoody.com Connect to Hoody

You can add what you want done in the same message:

@https://hoody.com Connect to Hoody, then put my website online

Any AI that can run code with network access can do this: Claude, coding agents such as Claude Code, Codex and Cursor, and other chat apps that offer it. Write in your own language: your AI answers in it.

  1. Your AI reads Hoody’s instructions. It opens hoody.com and finds a guide written for AI. Nothing is installed on your device.
  2. It asks what you want to do, unless you already said. It can also ask how much you want explained, so a first-timer and a developer both get the right pace.
  3. It signs you in through your browser. Your AI gives you a link and a short code. Open the link, check that the page shows the same code, then sign in and approve. If you’re new, it gives you the signup link instead: create your account, then click the link in the email Hoody sends you before you sign in. Eligible accounts normally receive a free server and a first computer after verification and sign-in. Setup may require an invite code or remain on hold. Pricing shows what it includes.
  4. It gets your computer ready. Your first computer is a full Linux machine that keeps running when you close the chat.
  5. It does the task, and gives you a link to see the result: your website, an app in a browser tab, a file.

Your AI should never ask for your password, and you don’t need to paste it into the chat: you type it on Hoody’s own sign-in page. If you didn’t ask your AI to sign you in, choose Don’t authorize this device on that page.

Your AI has to do two things: read a web page, and send requests to Hoody. Reading is enough to explain Hoody. Signing you in and creating anything takes requests.

Your AIWhat to do
Claude on the web, desktop or mobileTurn on code execution and network access (below)
Claude Code in your terminalNothing, unless you turned on its sandbox (below)
Claude Code on the webAllow Hoody in the environment’s network access (below)
Another chat app or coding agentIf it has a code or network setting, turn it on and allow Hoody (below)

If your AI can only read pages, it tells you so. It can still write the commands for you to run on your own computer.

Many AI apps run code in a sandbox that can only reach approved websites. A request to Hoody from there is refused until you add Hoody to the list. Add this one entry:

*.hoody.com

It covers the Hoody API and every computer address, such as https://{projectId}-{computerId}-files-1.{server}.computers.hoody.com. If an app has no place for a list, choose its option that allows all domains.

This is Claude at claude.ai, in your browser. The same settings apply to the Claude desktop and mobile apps.

  1. Open claude.ai/settings/capabilities. Or select your name in the sidebar, then Settings, then Capabilities.
  2. Turn on Code execution and file creation. Newer layouts call it Create and edit files.
  3. Turn on Allow network egress.
  4. Under Additional allowed domains, add *.hoody.com and save. If your settings show no domain list, or Hoody stays blocked, choose All domains instead.
  5. Start a new chat, and paste @https://hoody.com Connect to Hoody again. A chat that was already open keeps its old settings.

On a Team or Enterprise plan, only an owner or admin can change this, for everyone at once. They open claude.ai/admin-settings/capabilities (Organization settings, then Capabilities), turn on Code execution and file creation, choose Allow network egress to package managers and specific domains, and add *.hoody.com.

  1. Where you start a session at claude.ai/code, open the environment selector under the message box and edit your environment.
  2. Set Network access to Limited, and keep Allow package managers checked.
  3. In Allowed domains, add *.hoody.com.

Running sessions pick up the change within about a minute. On a Team or Enterprise plan, an owner edits shared environments on the Cloud environments admin page.

Claude Code asks before it runs a command or opens a website: approve its requests to Hoody. If you turned on its sandbox, add *.hoody.com to sandbox.network.allowedDomains in ~/.claude/settings.json:

{
"sandbox": {
"network": {
"allowedDomains": ["*.hoody.com"]
}
}
}
What you seeWhat it meansWhat to do
Host not in allowlist, or a 403 that names a host not allowedYour AI app blocked the request, not HoodyAdd *.hoody.com (above). In a Claude chat, start a new chat; Claude Code on the web picks it up within about a minute
Your AI says it can only read pagesIt can’t send requests from this appTurn on code execution with network access, or run the commands it writes on your own computer
The sign-in link expiredThe code is valid for a short timeAsk your AI to sign you in again
The page says your email address isn’t verified yetYou haven’t clicked the link in Hoody’s emailClick it (check your spam folder), then sign in again on the same page. If the page says the code expired, ask your AI for a new link
A call fails with 401From the Hoody API, the sign-in is missing or has expired. From a password-protected computer URL, its proxy username or password is missing or incorrectFor the Hoody API, ask your AI to sign you in again. For a protected computer URL, send the username and password set in Proxy permissions
The browser sign-in doesn’t workSome setups can’t finish itIn the Hub (hub.hoody.com), open Keys and tokens, press New key, enter a Name, then press Make key and Copy key (the secret is shown once). Paste it into the chat (Create an auth token)

Once you approve, your AI can do anything your account can do: create and delete computers, read your files and spend your wallet balance. Keep these habits:

  • Read the approval page. It shows the name your AI gave itself, marked as unverified. Approve only a sign-in you started.
  • Give a limited token for routine work. A token can be limited to one realm, a few permissions and an expiry date (Limit what a token can do).
  • End access you no longer need. A browser sign-in is a session. To end it, end that session or sign out everywhere (Sign out and end sessions). Ending a session doesn’t delete the API keys made while it was signed in, so check your keys too: disable or delete any key your AI made, and any you pasted into the chat (Disable, delete or rotate a token).

In the Hub (hub.hoody.com) you can do all of this without code. To make a limited key, open Keys and tokens, press New key, enter a Name, then set Works in, Choose rights and Ends, and press Make key. To end a sign-in, open Account, go to Sign-in sessions, and press End on that session or End all other sessions. To stop a key, find it under Keys and tokens > API keys and choose Pause or Delete….

In a coding agent you use often, install the Hoody Skill once. The agent then knows Hoody in every session without the opening line:

Terminal window
npx skills add https://hoody.com/SKILLS/SKILL.md

The Skill teaches your agent how Hoody works. It doesn’t sign you in: your agent still signs you in through the browser or uses a token you give it. Agent Skill Bundle lists every file it can read.