On-demand Linux computers powered by Firecracker microVMs are provided for AI agents. This self-hosted solution offers a secure, isolated, and ephemeral compute environment, enabling AI to interact with a full Linux desktop or shell, run code, and develop applications. It benefits users by providing disposable, powerful compute resources controlled by AI.

How It Works

The system utilizes Firecracker microVMs, ensuring kernel-level isolation distinct from containerization. Each machine boots rapidly from memory snapshots, supports resource capping, and defaults to self-destruction after a TTL, enhancing security and cost-efficiency. Persistent storage via S3-backed volumes is optional. The control plane, boringd, manages microVM lifecycle, with host setup streamlined via a single script.

Quick Start & Requirements

• Linux: Ubuntu 24.04 (x86_64/arm64) with /dev/kvm. Run git clone https://github.com/michaelshimeles/boring-computers && cd boring-computers && npm install && ./infra/setup.sh root@YOUR_BOX_IP (requires BORING_ANTHROPIC_KEY).
• macOS (M3+): Requires brew install lima. Run ./infra/local/setup-local.sh (requires BORING_ANTHROPIC_KEY).
• Web UI: Configure apps/web/.env and run npm run dev -w web.
• Prerequisites: KVM (Linux), brew (macOS), npm, git, go, and an Anthropic API key.
• Docs: boringcomputers.com (showcase).

Highlighted Details

• Full Linux desktop (VNC) or headless shell with preinstalled coding agents (Claude, Codex, Cursor, Pi, Node, Python, Git).
• AI-driven operation: interacts via screen or executes code, providing live URLs for applications.
• Supports file drag-and-drop and arbitrary port forwarding.
• Enables forking running machines in ~35ms for state replication.
• Offers persistent volumes (S3-backed) and an MCP server for AI agent integration.
• Includes an Effect-native TypeScript client (boring-computers-sdk).

Maintenance & Community

Contributions are welcome. No specific community channels or maintainer details are highlighted in the README.

Licensing & Compatibility

Licensed under Apache 2.0, permitting commercial use and integration into closed-source projects. Self-hosted with user-provided API keys.

Limitations & Caveats

Windows 11 via WSL2 support is planned but not yet implemented. The primary focus is on self-hosted deployments.