Self-host Cloudroom core
core-v0.1.82Rust runtime for AI coding agents in isolated sandboxes
- Stars
- 1 049
- Port
- 9840 (HTTP API)
- Requires
- Node.js, Rust
What is Cloudroom core?
Cloudroom core is a Rust service that runs AI coding agents in isolated Linux sandboxes on your own hardware. Each agent gets a separate sandbox with full admin rights inside but no access to other sandboxes or the host. The core manages agent execution, durable request queues, session history, and exposes an HTTP API for remote control. It runs as a systemd service on Ubuntu 24.04 (x86_64 or arm64), uses a Bearer token for authentication, and listens on localhost by default with Tailscale recommended for remote access. Agents use your own model subscriptions and credentials. The software is pre-release (pre-1.0) and licensed under Apache-2.0.
Should you self-host Cloudroom core?
Cloudroom core scores 50/100 on selfhost.directory (how scoring works) and ranks #96 of 209 in AI Agents. It has 1 049 GitHub stars.
It shipped 10 stable releases in the past year. The latest, core-v0.1.82, came out on 28 October 2026. New versions land roughly every 1 days.
Setup is advanced: plan for several services and some ongoing maintenance. It is Apache-2.0-licensed and written mostly in Rust.
Its closest alternative on selfhost.directory is Orca (91/100, 108 642 stars). It scores higher overall, so compare the two before you pick. See Cloudroom core vs Orca side by side.
Features
- Child threads / subagents
- Command Guard
- Desktop app integration
- Disk safety and pause/resume
- HTTP API with Bearer token auth
- In-place upgrades
- Isolated agent sandboxes
- Multiple AI agent support
- Session history and replay
- Tailscale remote access
Cloudroom core README
FAQ
How does Cloudroom keep my code and keys safe?
Each agent runs in its own isolated sandbox with full admin rights inside but no access to other users or sandboxes. Agents call model providers directly. API keys shared go through a secure form and never appear in chat.
Can I self-host Cloudroom?
The Rust core is designed to run independently on a Linux VM with PostgreSQL and supported harness credentials. Managed hosting is optional. Public release packaging and a polished self-hosting setup are still being prepared.
Will my agents work when my laptop is closed?
Yes, on Cloud threads. Once a task reaches the cloud, the agent keeps working and its history is saved. Local threads stop when your Mac sleeps.
Do I need to change my model subscription?
No, Cloudroom works with your own model accounts and coding subscriptions. Agents connect to model providers directly.
Which AI coding agents does it support?
Local threads run 9 harnesses: Claude Code, Codex, Pi Agent, Cursor, OpenCode, Grok, fx, Hermes, and Devin. Cloud threads run Claude Code, Codex, Pi Agent, OpenCode, fx, Hermes, and Devin.
Is Cloudroom free?
Yes, the Free plan includes 100 cloud agents every 30 days, and local agents are free forever. Paid plans (Pro $50/mo, Super $150/mo, Ultra $500/mo) give more capacity.
Can I use Cloudroom today?
Yes, you can create an account and download the app. Ultra teams get an onboarding call.
What is Cloudroom?
Cloudroom is a platform for running AI coding agents in the cloud, each in its own sandbox. It manages agent execution and session history while you use your own tools and model accounts.
More in AI Agents
CoderCoder is an open-source, self-hosted control plane for remote dev environments and AI coding agents, with Terraform-defined workspaces and WireGuard-secured access.
Codex MemoriesA self-hosted, open-source memory layer that lets OpenAI Codex remember project decisions locally via Git and SQLite, without any hosted vector database.
ColliePWASelf-hosted, open-source PWA that mirrors your Herdr, tmux or zellij agent panes on your phone, so Claude Code, Codex and OpenCode can be answered anywhere.
commonlyOpen-source, self-hosted team space where people and AI agents share pods with persistent memory, task boards, and Slack/Telegram connectors.
ContainariumSelf-hosted open-source agent runtime giving each AI a persistent SSH-reachable Linux box on LXC or Kubernetes with eBPF isolation and built-in MCP servers.
CORECORE is a self-hosted personal AI OS: a temporal memory graph, a task engine with approval, scratchpad, voice and messaging access, plus Claude Code and Codex sessions.
CotalCotal is an open-source, self-hosted pub/sub protocol letting AI agents from different providers coordinate in a shared space with presence and durable delivery.
crabwalkSelf-hosted, open-source companion monitor that streams OpenClaw (Clawdbot) agent activity into a live node graph, covering WhatsApp, Telegram, Discord and Slack.
crowSelf-hosted agentic framework and MCP platform to build AI agents, connect clients like Claude Code as native MCP connectors, with persistent memory and P2P sharing.
CubePlexA self-hosted AI workspace giving teams a persistent agent with shared context, approved tools, isolated execution, and chat platform connections.
DashClaw🛡️The governance runtime for AI agents. Intercept actions, enforce guard policies, require approvals, and produce audit-ready decision trails.
DeepSeek HarnessSelf-hosted, open-source (MIT) AI agent harness from DeepSeek: everything is a plugin. Run a local Web UI via npx, plus desktop apps, scheduled tasks, subagents and MCP.
0 comments