Quickstart 5-minute setup

1. Install the CLI

Install OpenClaw globally via npm or pnpm:

npm install -g @openclaw/mekong-cli
# or
pnpm add -g @openclaw/mekong-cli

Verify the installation:

mekong --version
# OpenClaw v5.0.0

2. Initialize Your Company

Run the init wizard to scaffold your company configuration:

mekong company/init

This creates .mekong/company.json with your company profile, active departments, and billing settings.

3. Run Your First Command

Cook a mission — the most powerful command in OpenClaw:

mekong cook "Write a product description for my SaaS"

The PEV engine will plan, execute, and verify the task. Output is saved to .mekong/outputs/.

Try other quick commands:

# Generate a pitch deck outline
mekong pitch "AI-powered inventory management for SMBs"

# Write a cold email sequence
mekong sales/email "B2B SaaS outreach, 5-email sequence"

# Create a sprint plan
mekong sprint "MVP launch — 2 weeks"

4. Configure Your LLM

OpenClaw supports any OpenAI-compatible API. Set three environment variables:

export LLM_BASE_URL=https://openrouter.ai/api/v1
export LLM_API_KEY=sk-or-v1-yourkey
export LLM_MODEL=anthropic/claude-sonnet-4

Or add them to .env in your project root. Pre-configured presets live in mekong/adapters/llm-providers.yaml.

Supported providers (auto-fallback chain):

• OpenRouter (recommended — access all models)
• Anthropic Claude directly
• OpenAI / Azure OpenAI
• Google Gemini via DashScope
• DeepSeek, Qwen
• Ollama (local, offline)

5. Deploy to Cloudflare

OpenClaw ships a zero-cost 3-layer CF infrastructure scaffold:

# Scaffold frontend + edge API
bash mekong/infra/scaffold.sh myproject startup

# Full stack (frontend + API + backend + D1 + KV + R2)
bash mekong/infra/scaffold.sh myproject scale

# Deploy
git push origin main  # CF Pages auto-deploys
wrangler deploy       # Workers backend