How to Run OpenClaw in Docker: Complete Setup Guide (2026)

Why Run OpenClaw in Docker?

Running OpenClaw in Docker gives you a reproducible, isolated environment that works identically across machines. No more "works on my machine" problems — your configuration, dependencies, and runtime are all packaged together.

Key benefits of the Docker approach:

• Consistent environment across development, staging, and production
• Easy updates — pull the latest image and restart
• Isolation — OpenClaw runs in its own container without affecting your host system
• Simple rollbacks — pin to a specific image tag to revert anytime
• Multi-instance support — run multiple OpenClaw instances with different configs

Prerequisites

Before you start, make sure you have:

• Docker Engine 24+ (or Docker Desktop 4.25+ on macOS/Windows)
• Docker Compose v2 (included with Docker Desktop; install separately on Linux)
• An API key from your LLM provider (OpenAI, Anthropic, or a local model URL)

Verify your Docker installation:

docker --version
docker compose version

Quick Start with Docker Run

The fastest way to get OpenClaw running in Docker:

docker run -d \
  --name openclaw \
  -p 3000:3000 \
  -e OPENCLAW_API_KEY=your-api-key-here \
  -e OPENCLAW_LLM_PROVIDER=anthropic \
  -v openclaw-config:/home/openclaw/.config/openclaw \
  -v openclaw-data:/home/openclaw/.local/share/openclaw \
  ghcr.io/openclaw/openclaw:latest

This starts OpenClaw with the web UI on port 3000. However, for production use, Docker Compose is the better approach.

Docker Compose Setup

Create a docker-compose.yml file in your project directory:

version: "3.8"

services:
  openclaw:
    image: ghcr.io/openclaw/openclaw:latest
    container_name: openclaw
    restart: unless-stopped
    ports:
      - "3000:3000"
    environment:
      - OPENCLAW_API_KEY=${OPENCLAW_API_KEY}
      - OPENCLAW_LLM_PROVIDER=${OPENCLAW_LLM_PROVIDER:-anthropic}
      - OPENCLAW_LOG_LEVEL=${OPENCLAW_LOG_LEVEL:-info}
      - OPENCLAW_MAX_CONTEXT=${OPENCLAW_MAX_CONTEXT:-8192}
    volumes:
      - ./config:/home/openclaw/.config/openclaw
      - ./data:/home/openclaw/.local/share/openclaw
      - ./skills:/home/openclaw/.local/share/openclaw/skills
    healthcheck:
      test: ["CMD", "curl", "-f", "http://localhost:3000/health"]
      interval: 30s
      timeout: 10s
      retries: 3

Create a .env file alongside it:

OPENCLAW_API_KEY=sk-your-key-here
OPENCLAW_LLM_PROVIDER=anthropic
OPENCLAW_LOG_LEVEL=info
OPENCLAW_MAX_CONTEXT=8192

Start the stack:

docker compose up -d

Volumes and Data Persistence

The three volume mounts ensure your data survives container restarts:

| Mount | Purpose | Contents | |-------|---------|----------| | ./config | Configuration | config.yaml, API keys, permissions | | ./data | Application data | Logs, conversation history, caches | | ./skills | Installed skills | Skill packages and manifests |

To back up your entire OpenClaw state, simply copy these three directories:

tar -czf openclaw-backup-$(date +%Y%m%d).tar.gz config/ data/ skills/

Platform-Specific Notes

macOS (Docker Desktop)

Docker Desktop on macOS uses a Linux VM under the hood. For best performance:

• Allocate at least 4GB RAM to Docker Desktop (Preferences → Resources)
• Enable VirtioFS for faster file sharing (Preferences → General)
• If using Apple Silicon, the ARM64 image is pulled automatically

# Check your architecture
docker inspect ghcr.io/openclaw/openclaw:latest | grep Architecture

Linux

Docker runs natively on Linux, giving you the best performance:

# Install Docker Engine (Ubuntu/Debian)
curl -fsSL https://get.docker.com | sh
sudo usermod -aG docker $USER

# Install Docker Compose plugin
sudo apt install docker-compose-plugin

File permissions matter on Linux. Make sure the mounted directories are writable:

mkdir -p config data skills
chmod 755 config data skills

Windows (Docker Desktop / WSL2)

On Windows, Docker Desktop with WSL2 backend is recommended:

• Enable WSL2 backend in Docker Desktop settings
• Store your docker-compose.yml inside the WSL2 filesystem for better I/O performance
• Avoid mounting Windows NTFS paths — use WSL2 paths instead

# Inside WSL2
cd ~/openclaw
docker compose up -d

Environment Variables Reference

| Variable | Default | Description | |----------|---------|-------------| | OPENCLAW_API_KEY | — | Your LLM provider API key (required) | | OPENCLAW_LLM_PROVIDER | anthropic | LLM provider: anthropic, openai, ollama | | OPENCLAW_LOG_LEVEL | info | Log verbosity: debug, info, warn, error | | OPENCLAW_MAX_CONTEXT | 8192 | Max context window tokens | | OPENCLAW_PORT | 3000 | Web UI port | | OPENCLAW_SKILL_DIR | /home/openclaw/.local/share/openclaw/skills | Custom skill directory | | OPENCLAW_DISABLE_TELEMETRY | false | Disable anonymous usage telemetry |

Running with Local Models (Ollama)

To use OpenClaw with Ollama for fully local AI:

version: "3.8"

services:
  openclaw:
    image: ghcr.io/openclaw/openclaw:latest
    container_name: openclaw
    restart: unless-stopped
    ports:
      - "3000:3000"
    environment:
      - OPENCLAW_LLM_PROVIDER=ollama
      - OPENCLAW_OLLAMA_URL=http://ollama:11434
      - OPENCLAW_OLLAMA_MODEL=llama3
    volumes:
      - ./config:/home/openclaw/.config/openclaw
      - ./data:/home/openclaw/.local/share/openclaw
    depends_on:
      - ollama

  ollama:
    image: ollama/ollama:latest
    container_name: ollama
    ports:
      - "11434:11434"
    volumes:
      - ollama-models:/root/.ollama

volumes:
  ollama-models:

Pull a model and start:

docker compose up -d
docker exec ollama ollama pull llama3

Updating OpenClaw

To update to the latest version:

docker compose pull
docker compose up -d

To pin a specific version for stability:

image: ghcr.io/openclaw/openclaw:v2.4.1

Troubleshooting

Container exits immediately

Check the logs:

docker compose logs openclaw

Common causes: missing API key, invalid config file, or port conflict.

Cannot connect to web UI

Verify the container is running and the port mapping is correct:

docker compose ps
curl http://localhost:3000/health

Skills not loading

Ensure the skills volume is mounted correctly and has the right permissions:

docker exec openclaw ls -la /home/openclaw/.local/share/openclaw/skills

High memory usage

Reduce the context window size and disable unused features:

OPENCLAW_MAX_CONTEXT=4096

Permission denied on mounted volumes (Linux)

Match the container user UID to your host user:

docker exec openclaw id
# Then on host:
sudo chown -R 1000:1000 config/ data/ skills/

Further Reading

• Self-Hosting OpenClaw: Docker Compose + Security Hardening Guide — Add reverse proxies, SSL, and security hardening
• Getting Started with OpenClaw — Native installation and initial configuration
• Is OpenClaw Safe? A Complete Security Guide — Understand the security model and best practices