Aii Cloud runs AI agents for you right in the browser. Now it can run them somewhere new — on a machine you own.
You still talk to the agent in your browser, but the actual work runs on the computer you connected. Give it a task from cloud.aiiware.com and it executes right there — reading your files, running your shell, using your toolchain — with a local approval gate on every action.
We call it cloud brains, local hands.
"Computer" is bigger than your laptop
The connected machine is anything you can run one command on:
- your dev laptop
- a VPS or bare-metal box
- a Docker container
- a Kubernetes pod
- a cloud instance (EC2, GCE, a droplet — whatever you have)
- the homelab server in the closet
In the demo above, the machine isn't a laptop at all — it's a remote Linux box (aiiware-dev). It connects with a single command, waits for work, and when a task lands from the browser, it runs it right there and reports back into the chat.
That's the real unlock: point the agent at the machine where your work already lives. The server that has your data. The container that has your exact toolchain pinned. The cloud box with the compute. You get a clean browser UI without giving up the environment that makes the work possible.
What the demo shows
From the browser I ask: "Find me the current status of @AiiCLI on Moltbook." The agent — running on the connected machine — uses that machine's tools to pull the live data and answers in the chat with the profile stats and recent activity. Nothing was pre-loaded; it went and got it.
Two details matter more than the answer:
It ran on the machine, not in the cloud. Every tool call shows Allowed · ran locally · in terminal. The file access and shell commands happened on the box I connected, in the working directory it reported at connect time (/home/aii/codespace/moltbook-aiicli) — not in a hidden container you can't see.
You stay in control. Actions are gated by the same local approval flow you'd get in the terminal, surfaced as cards in the browser. The agent proposes; you decide. Your own settings apply to the run.
How it works
Under the hood it's a doorbell, not a tunnel. Your machine runs a lightweight daemon that holds a connection to Aii Cloud and waits. When you dispatch a task from the Portal, the cloud rings that doorbell; the daemon wakes, opens a session, and runs the task on your machine — streaming the result back to the browser as it goes. Follow-up messages stay on the same machine, with context, for the whole conversation.
Local execution keeps the filesystem, shell, and credentials on your machine; Aii does not upload the computer wholesale. Prompts, assistant text, and tool results needed for a turn still travel through the conversation and inference path you selected, so keep secrets out of prompts and outputs. The machine acts only through the normal approval boundary.
Try it
On any machine you control — local or remote — install the CLI, sign in, and start an interactive session with remote control enabled:
npm i -g @aiiware/aii
aii cloud login
aii --remote-control --inference cloudOpen the AiiCloud link printed by the CLI to continue that conversation from the browser. While the process is running, its declared workspaces are also available under Run on ▾ when you start a new Work conversation.
--remote-control decides whether the browser can reach the session. The example uses AiiCloud inference; if you have configured an Aii Server, use --inference server instead. Remote control and inference are separate choices.
Run that session on a VPS and you've got an agent that works on your server while your laptop sleeps. Run it in a container and every action is sandboxed to that container's world. Run it on your dev box and the agent is already standing in your repo.
One browser. Any machine. Your rules.
Aii is the AI assistant for everything — terminal, editor, and browser. This is part of our "bring your own computer" work in Aii Cloud.