Supported operating systems
| Platform | Supported versions | Processors | Notes |
|---|---|---|---|
| macOS app | macOS 15 Sequoia or later | Apple silicon, Intel | Voice needs macOS 26+. Required for iMessage. |
| macOS CLI | Older macOS with supported Node.js | Apple silicon, Intel | No menu bar app |
| Windows | Windows 10 20H2+, Windows 11 | x64, ARM64 | Native, or inside WSL2 |
| Linux CLI | Most modern 64-bit distros | x64, ARM64 | Works headless |
| Linux desktop app | glibc 2.35+ (Ubuntu 22.04+, Debian 12+) | x64 only | Not RHEL 9 or Rocky 9 |
| Raspberry Pi | 64-bit Raspberry Pi OS | Pi 4, Pi 5 | 32-bit not supported |
| Docker | Docker Engine or Desktop with Compose v2 | Host-dependent | 6 GB RAM to build images |
| Phones | iOS and Android companion apps | — | Pair with a gateway; don't run one |
How much RAM, CPU and storage OpenClaw needs
With a cloud AI model, OpenClaw itself is light. The gateway routes messages and runs tools, and the heavy thinking happens on your model provider's servers.
One assistant, cloud model, a couple of chat apps.
Browser automation, several channels, multiple agents and room to grow.
Running AI models on your own hardware. Needs depend on model size.
These tiers are our recommendations from real-world setups. The official docs set two firm numbers: 6 GB of RAM to build the Docker image, and 2 GB minimum (4 GB recommended) on a Raspberry Pi. An SSD is strongly recommended over an SD card or spinning disk.
Software requirements
- Node.js
- 24.16+ or 26.1+, with 26 recommended. The installer script, local-prefix installer, desktop apps and Docker include it, so you only install it yourself for npm, pnpm or bun installs.
- npm
- npm 12, or 11.16+, for the
--allow-scriptsinstall flag. Older npm works without the flag. - pnpm
- Only needed to build from source
- Docker
- Docker Engine or Desktop with Compose v2 (
docker compose) - git
- For Docker setup and source installs
- Browser
- Any modern browser for the Control UI. Chromium on servers for the browser tool.
Check what you have:
node -v
npm -v
docker compose versionNetwork and account requirements
Outbound HTTPS to your AI model provider and chat services. No inbound ports are needed for most channels, including Telegram.
The gateway and Control UI listen here on 127.0.0.1. Keep it closed to the internet and use an SSH tunnel or Tailscale for remote access.
An API key from a provider such as Anthropic, OpenAI or Google, or a local model through Ollama, LM Studio or vLLM.
An account on each app you connect, for example a Telegram bot token or a WhatsApp number.
Features with extra requirements
| Feature | What you need |
|---|---|
| iMessage | A Mac signed in to Messages, the imsg tool, Full Disk Access and Automation permissions |
| Voice in the Mac app | macOS 26 or later |
| Talk mode on Linux | A regular web browser, because the desktop app window can't use the microphone |
| Browser automation | Chromium. On Docker, use the latest-browser image. |
| Camera, screen, location | The iOS or Android app, or Windows Hub, paired as a node |
| Sandboxed tools | Docker or Podman |
| 24/7 scheduled jobs | A machine that stays on, such as a VPS, Mac mini or Raspberry Pi |
| Global shortcuts on Linux | X11. On Wayland, use the tray's Quick Chat instead. |
Requirements for local AI models
Running the AI model on your own hardware keeps everything private and removes per-message costs, but the model needs far more resources than OpenClaw itself.
- Memory decides model size. As a rough guide, small models run in 8–16 GB, while larger, more capable models need 32 GB or more of RAM or GPU memory.
- Apple silicon Macs work well because the GPU shares system memory.
- On PCs and servers, a GPU with plenty of VRAM makes the biggest difference.
- Don't run local models on a Raspberry Pi. The official docs recommend cloud models there.
- Smaller models are less reliable at multi-step tool use than top cloud models.
Supported local runtimes include Ollama, LM Studio and vLLM. See local models with OpenClaw.
Recommended OpenClaw setups
Great for learning. Jobs pause when the laptop sleeps.
Install guideQuiet, efficient, and the only 24/7 option with iMessage.
Mac guideAlways on, low cost and separate from your personal computer.
VPS guideThe cheapest always-on option. Cloud models only.
Linux & Pi guideOpenClaw system requirements: questions
What are the minimum requirements for OpenClaw?
A 64-bit computer running macOS 15+, Windows 10 (20H2+) or 11, or a modern Linux distribution, with at least 2 GB of RAM, an internet connection and an AI model API key. 4 GB of RAM or more is recommended.
How much RAM does OpenClaw need?
With a cloud AI model, 2 GB is the practical minimum and 4 to 8 GB is comfortable. Building the Docker image needs 6 GB. Local AI models need much more, typically 16 GB or more.
What version of Node.js does OpenClaw need?
Node.js 24.16+ or 26.1+, with Node 26 recommended. The installer script and desktop apps set it up for you, so you only need your own Node.js for npm, pnpm or bun installs.
Does OpenClaw need a GPU?
No. With a cloud AI model, OpenClaw runs on an ordinary CPU. A GPU, or an Apple silicon Mac, only matters if you run AI models locally.
Can OpenClaw run on a Raspberry Pi?
Yes, on a Raspberry Pi 4 or 5 with a 64-bit OS and at least 2 GB of RAM (4 GB recommended). Use cloud AI models, and a USB SSD for best performance.
Does OpenClaw work on 32-bit systems?
No. OpenClaw requires a 64-bit operating system on every platform, including Raspberry Pi.