Why Cloud Terminal?

The problem we kept hitting

For two decades the working developer has lived inside a laptop. SSH out to a server when needed, but the workspace itself — editor, shell, history, scratch repos, half-built ideas — lives on a machine that you have to carry, charge, back up, and replace every three years. Close the lid mid-flow and you lose your tmux pane. Switch machines and you spend two days reconstructing dotfiles. Try to share a session with a teammate and you reach for Zoom screen-share like it is 2014.

Cloud IDEs solved one slice of this and re-introduced two: they are tied to one editor (theirs), they bury you in their UI chrome, and they treat the terminal like a second-class citizen. Most of them never gave you a real shell — they gave you a sandbox that pretends to be one. The moment you wanted to run sixteen Claude Code agents in parallel, attach to tmux from your phone, or sign into a different account per session, you bounced off the walls of someone else's product.

What we built instead

cloudscode is a real terminal — bash, zsh, tmux, your dotfiles, your whole shell — running on a per-tenant container in your chosen region, reachable from any browser. It boots in seconds, persists forever, and never tries to wrap your workflow in someone else's UI. Open a tab, you are in. Close the tab, the session keeps running. Open it on your phone, same prompt. Hand the URL to a teammate, they pair-program with you live.

On top of that we layered something nobody else ships: up to sixteen AI agents in parallel, each in its own pane, each signed into its own account if you want, each watching the same filesystem. Bring Claude, Codex, Gemini, OpenClaw — or none. The multi-agent layer is optional; the terminal underneath is the product. We do not gate one behind the other.

The 16-agent thesis

A single agent is a chat window. Sixteen agents is a workforce. One handles the migration, one writes the tests, one ships the staging branch, one reviews the diff, one wires up the registry, one runs the smoke tests, one reads the logs and reports back. They share a worktree, they coordinate over branches, and they finish in parallel what a single human-and-agent pair would serialize. We run our own internal pipeline this way every day — sixteen Opus agents live, hard target, always — and it is not a demo. It is how the product gets built.

Inherited compliance, honest posture

We do not yet hold our own SOC 2. We say so on our compliance page, in plain English, before you ask. What we DO hold is a small, audited supply chain. Cloudflare and Hetzner are both SOC 2 Type II + ISO 27001 + GDPR. Every byte of your data lives behind those certifications. Region pinning is real (Singapore, EU, or US, your choice), per-tenant containers are real (named volumes nobody else can mount), and BYOK keys are envelope-encrypted at rest and never logged. We will pursue our own SOC 2 when the customer base demands it; until then we are honest about what we inherit and what we do not.