How-to4 min read
How to run Claude Code in a browser, on any device
The agent CLI is already installed and running on the machine — you reach it through the same tab as everything else.
Claude Code is a terminal program. That is the whole reason it works so well here: your machine has a terminal, reachable from any browser, so the agent runs on the box and you watch it from wherever you happen to be.
It is already installed
You do not install it, and you do not open a port for it. It is Anthropic's own CLI — anthropics/claude-code, from the package they publish, not a fork or a wrapper. Open the terminal tile, change into a project, and start it:
cd ~/my-project
claudeFrom there it is the CLI you already know — it reads and edits files in that directory, runs commands, and asks before doing anything it should ask about.
Why the browser is enough
An agent that edits code wants to be next to the code, on a machine with the toolchain installed. Running it locally means keeping that toolchain on your laptop; running it here means the laptop only has to draw a terminal. The heavy part — the repository, the language servers, the test runner — is already on the VM.
Sessions survive the tab
Start the agent inside the persistent workspace rather than a bare terminal and it keeps running when you disconnect. Kick off a long refactor, close the tab, come back in an hour and read what it did. The session did not end because your browser did.
More posts
How-to4 min read
How to keep an AI agent running 24/7 after you close the tab
Detach from a session instead of ending it, and a long job keeps running while you are gone.
How-to5 min read
Running Docker on a cloud dev machine, with root on a real VM
You have root on a real VM, which means Docker runs the way it does anywhere else.
How-to5 min read
How to point your own domain at your VM, with a real TLS certificate
It is a real VM at a real provider, so a domain and a real certificate are a DNS record away.