Home Installation & Setup System Requirements System Requirements
System Requirements Pre-Install Hardware

OpenClaw System Requirements: 5 Things You Must Check First

Checking requirements takes 60 seconds. Recovering from a broken install on an unsupported system takes hours. Do it first.

MK
M. Kim
Platform Engineer
Jan 17, 2025 11 min read 7,600 views
Updated Jan 24, 2025
Key Takeaways
Node.js 18+ is the single most-missed requirement — check this first on every machine
2GB RAM minimum, 4GB recommended for multi-agent use — RAM is the bottleneck for parallel agents
No GPU required — OpenClaw connects to cloud models via API, not local inference
Supported: macOS 11+, Ubuntu 20.04+, Debian 11+, Alpine 3.15+, NixOS, Windows 10/11 via WSL2
Network access to model provider API endpoints is required — air-gapped setup needs local inference

Three out of four failed OpenClaw installs we've seen come down to one of two causes: wrong Node.js version or insufficient RAM for the user's intended workload. Both are easy to check before installing. Both are invisible in the official quick-start docs. Check all five requirements below before you touch the install command — it saves the debugging session you'd otherwise spend tomorrow.

Requirement 1: Node.js Version

OpenClaw is a Node.js application. It will not run on Node 16 or lower. Period.

node --version
# Must return v18.x.x or higher

If you get v16 or lower, upgrade before proceeding. The install command will appear to work, then fail silently at runtime with cryptic module errors that look like bugs rather than a version mismatch.

Node version recommendations as of early 2025:

  • Node 20 LTS — recommended for production; stable, well-tested with OpenClaw
  • Node 22 — works but is newer and less tested; suitable for development
  • Node 18 LTS — minimum supported; works but misses some performance improvements
  • Node 16 or lower — not supported, install will fail

Requirement 2: RAM

OpenClaw uses approximately 150–300MB of RAM per running agent instance. That's the runtime itself — not counting the Node.js overhead (~80MB), OS baseline, or any other processes.

Practical minimums:

  • Single agent, basic use — 2GB total system RAM
  • Single agent with active channels — 4GB recommended
  • 3+ concurrent agents — 8GB recommended
  • Production server — 16GB for headroom during traffic spikes

The 2GB machines we've tested (certain Raspberry Pi models, low-end VPS instances) do run OpenClaw, but they swap under load. Swapping kills agent response latency. For anything beyond a personal project, 4GB is the real minimum.

⚠️
Memory Leaks in Long-Running Sessions
As of early 2025, there is a known gradual memory growth in sessions running longer than 72 hours with high message throughput. This is tracked in the OpenClaw issue tracker. If you run persistent agents, plan for weekly restarts or monitor RSS memory and set up auto-restart via PM2 or systemd when usage exceeds a threshold.

Requirement 3: Disk Space

The OpenClaw binary and its Node.js dependencies require approximately 200MB. Add to this:

  • Skills and plugins: 50–500MB depending on installed skills
  • Log files: grows with usage; default rotation keeps last 7 days
  • Config and cache: ~50MB typical

For a development machine, 1GB free disk space is comfortable. For a production server with logging enabled, plan for 5–10GB to avoid log rotation failures during high-activity periods.

Requirement 4: Operating System Support

Supported platforms as of early 2025:

Platform Minimum Version Support Level
macOS IntelmacOS 11 Big SurFull support
macOS Apple SiliconmacOS 12 MontereyFull support (native arm64)
Ubuntu20.04 LTSFull support
DebianDebian 11 BullseyeFull support
Alpine LinuxAlpine 3.15Supported (musl libc)
NixOSNixOS 22.11Supported
Windows (WSL2)Windows 10 21H2Full support via WSL2
Windows (native)Windows 10Limited — WSL2 preferred

Requirement 5: Network Access

OpenClaw connects to your model provider's API endpoints over HTTPS. The specific endpoints depend on your chosen provider:

  • OpenAI: api.openai.com (port 443)
  • Anthropic: api.anthropic.com (port 443)
  • OpenClaw update servers: get.openclaw.io (port 443)

Corporate firewalls that use SSL inspection can break these connections. If you're on a restricted network, check with your IT team about whitelisting these domains before troubleshooting the install itself.

💡
Local-Only Mode is Possible
If you need air-gapped operation, set up Ollama (ollama.ai) as a local model provider first. OpenClaw can connect to Ollama on localhost:11434, which means no external API calls after setup. You still need internet for initial install and updates.

Quick Requirements Check Script

Run this to check all five requirements in one pass:

#!/bin/bash
echo "=== OpenClaw Requirements Check ==="

# Node.js version
NODE_VER=$(node -e "process.exit(parseInt(process.versions.node.split('.')[0]) >= 18 ? 0 : 1)" 2>/dev/null && echo "OK" || echo "FAIL — need Node 18+")
echo "Node.js: $NODE_VER"

# RAM check
RAM_GB=$(free -g 2>/dev/null | awk '/^Mem:/{print $2}' || sysctl hw.memsize 2>/dev/null | awk '{printf "%.0f", $2/1073741824}')
echo "RAM: ${RAM_GB}GB (minimum: 2GB)"

# Disk space
DISK_FREE=$(df -BG ~/.  2>/dev/null | awk 'NR==2{gsub(/G/,"",$4); print $4}')
echo "Free disk: ${DISK_FREE}GB (minimum: 1GB)"

# OS
echo "OS: $(uname -s) $(uname -r)"

# Network
curl -s --connect-timeout 5 https://api.openai.com > /dev/null 2>&1 && echo "Network: OK" || echo "Network: Cannot reach api.openai.com"

Production Hardware Recommendations

For teams deploying OpenClaw as a shared service rather than a personal tool:

  • CPU — 4 cores minimum; OpenClaw is largely I/O-bound on API calls but benefits from multiple cores for parallel skill execution
  • RAM — 16GB for up to 10 concurrent agents with active channel connections
  • Storage — SSD required for good performance; HDD causes significant latency in skill loading and log writes
  • Network — 100Mbps+ for high-volume message processing; latency to model provider endpoints matters more than raw bandwidth

Frequently Asked Questions

What are the minimum system requirements for OpenClaw?

OpenClaw requires Node.js 18 or higher, 2GB RAM minimum (4GB recommended), 500MB disk space, and macOS 11+, Ubuntu 20.04+, Debian 11+, or Windows 10 with WSL2. Network access to your model provider's API endpoints is also required for cloud models.

Does OpenClaw require a GPU?

No. OpenClaw connects to cloud model providers via API, so no GPU is needed on the host machine. GPU acceleration only applies if you run a local inference server like Ollama alongside OpenClaw, which OpenClaw can connect to as a model provider for offline operation.

How much RAM does OpenClaw need?

OpenClaw uses 150–300MB RAM per agent instance. A single agent needs 2GB total system RAM minimum. For 3+ concurrent agents with active channel connections, 8GB is recommended. Memory usage scales with active skills and connected channels during peak load.

What operating systems does OpenClaw support?

macOS 11 Big Sur and later (Intel and Apple Silicon), Ubuntu 20.04+, Debian 11+, Alpine Linux 3.15+, NixOS, Arch Linux, and Windows 10/11 via WSL2. Native Windows support exists but WSL2 delivers significantly better performance for agent workloads.

What Node.js version does OpenClaw require?

Node.js 18.0.0 or higher is required. Node 20 LTS is recommended for production. Node 22 is supported but less tested as of early 2025. The install fails with an engines mismatch error on Node 16 or lower — upgrade Node before attempting installation.

Does OpenClaw require an internet connection?

The OpenClaw runtime can operate offline with local model providers like Ollama. Cloud-based providers (OpenAI, Anthropic) require internet access. Initial setup and updates also require internet connectivity. Air-gapped deployments are possible with local inference configured beforehand.

How much disk space does OpenClaw need?

The binary and dependencies require approximately 200MB. Config, logs, and skills add 100–500MB. For production deployments with extensive logging, plan for 2–5GB to accommodate log rotation and skill cache without disk space warnings appearing.

MK
M. Kim
Platform Engineer · aiagentsguides.com

M. Kim has benchmarked OpenClaw across hardware ranging from Raspberry Pi to 64-core bare-metal servers. She focuses on production readiness, capacity planning, and performance bottleneck identification for teams scaling OpenClaw deployments.

Pre-Install Checklist
System checks and capacity planning tips delivered weekly.