openclaw-installer 1.0.14

Zero-config installer for OpenClaw AI assistant - works with Kimi, Claude, GPT, and local models

🦞 OpenClaw Installer

Zero-config installer for OpenClaw AI assistant

Install and configure OpenClaw — the AI that actually does things — with a single command. Works with Kimi (Moonshot AI), Claude, GPT-4, OpenRouter, and local Ollama models.

---

✨ Features

• One-command install — No manual Node.js setup
• Multiple AI providers — Kimi, Claude, GPT-4, OpenRouter, Ollama
• Cross-platform — Windows, macOS, Linux
• Bundled Node.js — Uses nodejs-bin, no external dependencies
• Simple YAML config — One file, zero coding
• Auto-starts gateway — Ready to use immediately

---

🚀 Quick Start

Windows Users: Run PowerShell as Administrator (required for npm global installs)

1. Install pipx (if not already installed)

pipx runs the installer in an isolated environment, avoiding dependency conflicts with other Python packages.

Install pipx:

• macOS/Linux: https://pipx.pypa.io/stable/installation/#installing-pipx
• Windows: https://pipx.pypa.io/stable/installation/#install-on-windows

Or install via package manager:

# macOS
brew install pipx
pipx ensurepath

# Linux (Debian/Ubuntu)
sudo apt install pipx
pipx ensurepath

# Linux (Fedora)
sudo dnf install pipx
pipx ensurepath

# Windows (requires Python 3.8+)
python -m pip install --user pipx
python -m pipx ensurepath

2. Install OpenClaw Installer

pipx install openclaw-installer

Note: If you see a PATH warning like:

⚠️  Note: '/Users/username/.local/bin' is not on your PATH environment variable.

Run this to fix it:

pipx ensurepath

Then restart your terminal or run:

source ~/.zshrc   # For macOS Catalina and later (zsh)
# or
source ~/.bashrc  # For older macOS or if using bash

2. Configure

Create openclaw.yaml:

ai:
  provider: anthropic  # or openai, kimi, ollama
  key: "your-api-key"

channels:
  web: true

gateway:
  autostart: true

3. Install OpenClaw

# Linux/Mac
openclaw-installer install --config openclaw.yaml

# Windows (PowerShell as Administrator)
openclaw-installer install --config openclaw.yaml
```## 1. Install pipx (if not already installed)

See [Quick Start](#1-install-pipx-if-not-already-installed) above for pipx installation instructions.

## 2. Install OpenClaw Installer

```bash
pipx install openclaw-installer

2. Generate a config

openclaw-installer generate-config > openclaw.yaml

3. Edit the config

ai:
  provider: kimi           # or anthropic, openai, ollama
  key: "sk-your-key-here"  # Get from https://platform.moonshot.cn/

channels:
  web: true

gateway:
  autostart: true

4. Install OpenClaw

openclaw-installer install --config openclaw.yaml

5. Open Dashboard

openclaw dashboard
# or visit: http://localhost:18789

Done! 🎉

---

⚡ vs Official OpenClaw CLI

Why use this installer instead of npm install -g openclaw && openclaw onboard?

Feature	Official Wizard	This Installer
Node.js required?	✅ Yes (install manually)	❌ No (bundled automatically)
Installation style	Interactive prompts	YAML config (unattended)
CI/CD automation	❌ Manual only	✅ Fully automated
Repeatable deploys	❌ Answer prompts each time	✅ Idempotent
Non-technical users	Must understand Node/npm	✅ Edit a file, run a command
Version control	❌ Interactive state	✅ Config in git
Dependency isolation	❌ pip (potential conflicts)	✅ pipx (isolated env)

When to Use This Installer

✅ Perfect for:

• Teams sharing standard setups
• Docker containers / cloud VMs
• CI/CD pipelines
• Non-technical users who just want it to work
• Repeatable deployments across multiple machines

✅ Example scenario:

# Deploy to 50 company laptops
# 1. Share openclaw.yaml with API key
# 2. One command: pipx install openclaw-installer && openclaw-installer install --config openclaw.yaml
# 3. Everyone has identical setup

When to Use Official Wizard Instead

⚠️ Better choice when:

• Exploring OpenClaw for the first time
• Need hand-holding through provider selection
• Already have Node.js installed and prefer interactive setup
• Want to customize every option manually

Community Context

This package fills a specific gap:

• Available on PyPI: Package name openclaw-installer is reserved and ready
• Community demand: Existing GitHub projects like miaoxworld/OpenClawInstaller and Personal-MT-Engineering/openClaw-install show users want easier installation
• Different from armorer: That tool focuses on security sandboxing; this focuses on ease-of-use

Bottom line: After this installer runs, you use the same OpenClaw dashboard/CLI as everyone else. We're just the "easy on-ramp."

---

🖥️ Using the OpenClaw Dashboard

The Dashboard (Control UI) is your central hub for managing OpenClaw.

Opening the Dashboard

Method	Command
CLI	openclaw dashboard
Auto-Login	openclaw-installer auto-login
Browser	http://localhost:18789
Auto-open	Enabled by default after install

Auto-Login (Zero-Interaction)

The auto-login feature automatically connects to the dashboard without requiring manual token entry. It follows the "How to connect" steps:

1. Start the gateway on your host machine
2. Get a tokenized dashboard URL
3. Open the browser with the authenticated URL

# Auto-login with default browser
openclaw-installer auto-login

# Auto-login with specific browser
openclaw-installer auto-login --browser "/path/to/chrome"

# Open dashboard with auto-login (same as above)
openclaw-installer dashboard

Cross-Platform Support:

• ✅ Windows: PowerShell script + Python automation
• ✅ macOS: Bash script + Python automation
• ✅ Linux: Bash script + Python automation

Configure in your YAML:

automation:
  auto_open_dashboard: true  # Auto-open after install
  auto_login: true           # Skip manual token entry
  browser_path: ""           # Use system default
  startup_timeout: 30        # Seconds to wait for gateway

Dashboard Sections

🤖 Agents Panel

• Create Agents: Build specialized AI agents for different tasks
• Configure Memory: Set context windows and persistence
• Agent Teams: Group agents for collaborative workflows
• Skills: Install/enable capabilities (web search, code execution, etc.)

💬 Conversations

• Chat with your AI across all connected channels
• View conversation history
• Export/backup important chats

🔌 Integrations (Channels)

• Web UI: Always available at http://localhost:18789
• WhatsApp: Pair by scanning QR code
• Telegram: Configure bot token
• Discord: Add bot to servers
• Slack: Install workspace app
• iMessage: macOS only (requires Full Disk Access)

⚙️ Settings

• AI Provider: Switch between Claude, GPT, Kimi, Ollama
• Model Selection: Choose specific models per task
• System Prompts: Customize agent behavior
• Memory: Configure long-term storage
• Gateway: Port, logging, security settings

First-Time Setup in Dashboard

1. Verify AI Connection: Settings → AI → Test Connection
2. Enable Skills: Agents → Skills → Install (see below)
3. Create First Agent: Agents → New → Select skills
4. Test Chat: Conversations → New → Send test message

---

🤖 Supported AI Providers

Provider	Model Example	Setup	Best For
Kimi (Moonshot AI)	kimi-k2.5	platform.moonshot.cn	Chinese/English, 256K context
Anthropic	claude-3-5-sonnet	console.anthropic.com	Coding, reasoning
OpenAI	gpt-4o	platform.openai.com	General tasks
OpenRouter	anthropic/claude-3.5-sonnet	openrouter.ai	Model flexibility
Ollama	llama3.2	ollama.ai	Privacy, offline

---

📁 Example Configurations

Kimi (Moonshot AI)

ai:
  provider: kimi
  key: "sk-your-moonshot-key"
  model: "kimi-k2.5"

channels:
  web: true
  telegram: true
  telegram_token: "${TELEGRAM_BOT_TOKEN}"

Claude with all channels

ai:
  provider: anthropic
  key: "${ANTHROPIC_API_KEY}"
  model: "claude-3-5-sonnet"

channels:
  web: true
  whatsapp: true
  telegram: true
  telegram_token: "your-bot-token"

Local Ollama (offline)

ai:
  provider: ollama
  key: ""  # No key needed
  model: "llama3.2"

gateway:
  port: 18789

---

🛠️ Commands

Command	Description
openclaw-installer install --config file.yaml	Install OpenClaw from config
openclaw-installer generate-config	Print example config
openclaw-installer generate-config -o config.yaml	Save example config
openclaw-installer uninstall	Remove OpenClaw
openclaw-installer auto-login	Auto-login to dashboard (zero-interaction)
openclaw-installer dashboard	Open web dashboard with auto-login
openclaw dashboard	Open web dashboard (standard)
openclaw gateway status	Check if gateway running
openclaw daemon start/stop	Control background service
openclaw logs	View logs
--dry-run	Simulate without changes

Upgrading

pipx upgrade openclaw-installer

Uninstalling

pipx uninstall openclaw-installer

---

🛡️ Best Practices

🔐 Security

API Key Management

# ✅ GOOD: Use environment variables
ai:
  key: "${ANTHROPIC_API_KEY}"  # Reads from env var

# ❌ BAD: Hardcoded key
ai:
  key: "sk-ant-api03-actual-key-here"

Set environment variables:

# Linux/macOS
export ANTHROPIC_API_KEY="sk-ant-..."
export KIMI_API_KEY="sk-..."

# Windows PowerShell
$env:ANTHROPIC_API_KEY="sk-ant-..."

# Windows CMD
set ANTHROPIC_API_KEY=sk-ant-...

Gateway Security

• Default: Runs on 127.0.0.1 (localhost only) — SAFE
• Remote access: Set host: 0.0.0.0 only with authentication
• Firewall: Block port 18789 from external networks unless needed

# Local only (secure default)
gateway:
  host: 127.0.0.1
  port: 18789

# Remote access (use with caution)
gateway:
  host: 0.0.0.0  # Accessible from network
  port: 18789
  # Add reverse proxy with auth (nginx/traefik)

File System Permissions

# Workspace should be user-readable only
chmod 700 ~/openclaw-workspace

# Config files with API keys
chmod 600 openclaw.yaml

---

🧩 Installing & Managing Skills

Skills extend OpenClaw capabilities (web search, code execution, file operations).

Via Dashboard (Recommended)

1. Open Dashboard: openclaw dashboard
2. Navigate: Agents → Skills
3. Browse skill marketplace
4. Click Install on desired skills
5. Configure skill settings

Via CLI

# List available skills
openclaw skills list

# Install a skill
openclaw skills install web-search
openclaw skills install code-execution
openclaw skills install file-manager

# Enable/disable
openclaw skills enable web-search
openclaw skills disable code-execution

Essential Skills

Skill	Purpose	Install
web-search	Search Google/DuckDuckGo	openclaw skills install web-search
code-execution	Run code safely	openclaw skills install code-execution
file-manager	Read/write files	Built-in
browser	Web browsing	openclaw skills install browser
memory	Long-term memory	Built-in

Custom Skills

Create custom skills in your workspace:

mkdir -p ~/openclaw-workspace/skills/my-skill
cd ~/openclaw-workspace/skills/my-skill

Create skill.yaml:

name: my-custom-skill
description: Does something awesome
version: 1.0.0
entry: main.py

Create main.py:

def run(command: str) -> str:
    """Execute the skill."""
    return f"Executed: {command}"

---

👥 Creating Agent Teams

Agent teams let multiple specialized agents collaborate on complex tasks.

Dashboard Method

1. Agents → Teams → Create Team

2. Add agents with specific roles:

   • Researcher: Web search, data gathering
   • Coder: Code execution, debugging
   • Writer: Documentation, summaries
   • Reviewer: Quality checks

3. Set Orchestrator: The lead agent that delegates tasks

Example: Development Team

# Team configuration (in Dashboard)
team:
  name: "Dev Team"
  orchestrator: "tech-lead"
  agents:
    - name: "tech-lead"
      model: "claude-3-5-sonnet"
      skills: ["planning", "architecture"]

    - name: "code-reviewer"
      model: "claude-3-5-sonnet"
      skills: ["code-review", "security-audit"]

    - name: "researcher"
      model: "gpt-4o"
      skills: ["web-search", "documentation"]

CLI Method

# Create team
openclaw teams create dev-team --orchestrator tech-lead

# Add agents
openclaw teams add dev-team tech-lead --model claude-3-5-sonnet
openclaw teams add dev-team code-reviewer --model claude-3-5-sonnet
openclaw teams add dev-team researcher --model gpt-4o

# Activate team
openclaw teams use dev-team

---

🔗 Nodes (Distributed OpenClaw)

Nodes allow running OpenClaw across multiple machines for load distribution or specialized hardware.

Use Cases

• GPU Node: Remote machine with GPU for local AI (Ollama)
• Cloud Node: VPS running 24/7 for webhooks
• Home Node: Raspberry Pi for always-on gateway

Setting Up a Node

On Node Machine:

# Install OpenClaw with node role
openclaw-installer install --config node-config.yaml

# Expose node (behind firewall/VPN recommended)
openclaw node serve --port 18790 --token SECURE_TOKEN

On Main Machine:

# main-config.yaml
nodes:
  - name: "gpu-server"
    url: "http://192.168.1.100:18790"
    token: "SECURE_TOKEN"
    capabilities: ["ollama", "gpu"]

  - name: "cloud-vps"
    url: "https://vps.example.com:18790"
    token: "SECURE_TOKEN"
    capabilities: ["webhook", "always-on"]

# Route specific tasks to nodes
routing:
  ollama-requests: "gpu-server"
  webhook-handlers: "cloud-vps"

Node Commands

# List connected nodes
openclaw nodes list

# Check node health
openclaw nodes status gpu-server

# Send command to specific node
openclaw nodes exec gpu-server "ollama list"

---

💾 Backup & Migration

What to Backup

# Configuration
cp openclaw.yaml ~/backups/

# Workspace (agents, memory, skills)
tar -czf openclaw-backup.tar.gz ~/openclaw-workspace/

# Environment variables
env | grep OPENCLAW > openclaw-env.txt

Restore

# Restore workspace
tar -xzf openclaw-backup.tar.gz -C ~/

# Reinstall OpenClaw
openclaw-installer install --config openclaw.yaml

---

🚀 Performance Optimization

For Fast Response

ai:
  provider: anthropic
  model: "claude-3-5-haiku"  # Faster than sonnet

For Complex Tasks

ai:
  provider: anthropic  
  model: "claude-3-opus"  # Best reasoning

Local AI (Privacy + Speed)

ai:
  provider: ollama
  model: "llama3.2"  # Runs offline, no latency

---

🔧 Troubleshooting

Dashboard Won't Load

# Check gateway status
openclaw gateway status

# Restart gateway
openclaw daemon restart

# Check logs
openclaw logs | tail -50

API Key Not Working

# Test directly
export OPENCLAW_API_KEY="your-key"
openclaw config set ai.apiKey "$OPENCLAW_API_KEY"
openclaw config test-connection

Reset Everything

# Uninstall and clean
openclaw-installer uninstall
rm -rf ~/openclaw-workspace

# Fresh install
openclaw-installer install --config openclaw.yaml

---

📋 Requirements

• Python 3.9+
• No Node.js required (bundled)
• No npm required (bundled)

---

📄 License

MIT License

---

Not affiliated with OpenClaw. This is a community installer.