engineering / agent

Cloudroom

A desktop app for running Claude Code, Codex and Pi agents locally or in the cloud, with threads you can move between the two.

At a glance

Purpose
A desktop app for running Claude Code, Codex and Pi agents locally or in the cloud, with threads you can move between the two.
Runs in
Local agent with shell, Web service
Operating systems
macOS, Linux
Needs
Runtime: Not stated; Needs internet: Yes; Needs accounts or API keys beyond your agent: No
Files and data
Reads files: Yes; Writes files: Not stated; Sends data to other services: Yes
Tool type
Agent
Delivery
From GitHub
Version
Latest on the main branch
Access
Code on GitHub
Compatibility
According to the README, the Mac app is for Apple Silicon on macOS 13 or later, and a Linux x64 alpha is on Releases. The README says it runs Claude Code, Codex and Pi, and that agents call Claude and OpenAI directly on your own subscriptions. The README mentions login links for Codex and Cursor. Cloud agents run next to Cloudroom core, which you can self-host on a Linux machine you own or use through the project's hosting.
External costs
The README says local agents are free and need no account. Cloud is invite-only for now. It says agents call Claude and OpenAI directly on your own subscriptions, and that you can self-host Cloudroom core or use the project's hosting. It gives no prices for hosting or for those subscriptions.
Permissions
The README's source setup installs dependencies with pnpm and requires dependency install scripts for native SQLite, file-watching and terminal modules. The repository's top-level file list includes scripts/, but the README does not describe that folder. The README requires installed, authenticated provider CLIs for local agents and describes agents using your Claude and OpenAI subscriptions. The README says local agents run on the Mac, cloud agents receive repository copies in their sandboxes, and cloud agents can reach the Mac when allowed. The README describes computer use on the Mac and in the cloud. The README says cloud features require sign-in and manual core connections use a URL and token. The README describes storing the core token in the app's private data folder, restricting remote core connections to HTTPS and controls on local-API access. The README points to the core's security model for cloud execution.
Usage terms
Apache License 2.0, as stated in the repository. See its LICENSE file for the exact terms. The README adds that BB's MIT license and copyright notice are kept in the NOTICE file, and that bundled third-party notices remain with their components.
Interface
desktop

Code on GitHub

davidondrej/cloudroom-gui

Apache-2.0

TypeScript15 stars7 forksLast pushed 12 Oct 2026 UTC

Desktop app for Claude Code, Codex, and Pi. Start an agent on your Mac and move it to the cloud mid-task.

Topics: ai-agents background-agents claude-code cloud-agents codex coding-agents computer-use desktop-app

Public GitHub data fetched · 0 days old

View on GitHub ↗

What it does

Cloudroom is a desktop app for coding agents. The README describes running Claude Code, Codex and Pi on a Mac or in the cloud using your harnesses and subscriptions.

The README describes moving a running thread between Mac and cloud, per-agent cloud sandboxes with repository copies, direct Claude and OpenAI calls, permitted access back to the Mac and computer use. The README says cloud agents run alongside Cloudroom core, which can be self-hosted or accessed through the project's hosting.

The README says local agents are free and need no Cloudroom account, while cloud access is invite-only. The README identifies BB as the underlying project and says its MIT notice is retained in NOTICE.

Who it is for

People who use Claude Code, Codex or Pi and want one Mac app that can run them locally or in cloud sandboxes.

Example tasks

Run Claude Code, Codex and Pi side by side on your Mac.
Move a running thread from your Mac to the cloud and back.
Give a cloud agent its own sandbox with a copy of your repo.
Let a cloud agent reach your Mac when you allow it.
Build the app from source on macOS with pnpm.

Getting started

The README points to a download page for the Mac app (Apple Silicon, macOS 13 or later) and says cloud is invite-only for now. The README requires the same Node version for installation and builds. To build from source on macOS with Apple Silicon, the README gives pnpm install --frozen-lockfile and then pnpm exec turbo run build typecheck --filter=@bb/app --filter=@bb/server --filter=@bb/desktop --concurrency=2. To run locally, start pnpm dev, then in another terminal run pnpm exec turbo run dev --filter=@bb/desktop. The README says local agents need their own installed, authenticated provider CLIs. The README has the exact commands.

Limitations

According to the README, local agents need their own installed, authenticated provider CLIs. Cloud is invite-only for now with a waitlist, and cloud features need a signed-in account. Cloud execution needs a matching Cloudroom core with the direct_workspaces and command_guard capabilities, and the README names the public core release source-2026-09-23 or newer. The Mac app is for Apple Silicon on macOS 13 or later, and a Linux x64 alpha is on Releases. For building from source, the README gives macOS on Apple Silicon as the supported development target and lists Git, Node.js 22.19 or later, pnpm 9.15.0, Python 3.11 or later and Xcode Command Line Tools. The README says the build creates runnable bundles, not a signed installer. The repository can change at any time; check its README before you install.

Data practices

The README says agents call Claude and OpenAI directly on your own subscriptions, and that you keep your harness, subscriptions and traces. Local agents need no account. Cloud agents run in sandboxes next to Cloudroom core, which you can self-host or use through the project's hosting, and cloud features need a signed-in account. The README says the core token stays in the app's private data folder. It does not describe telemetry.

Community experience

Not rated yet

Sign in to share your experience.

Comments 0

No comments yet.

Sign in to comment

Report a listing concern