The best-in-class deployment tool for OpenClaw agent infrastructure.
One command. Full stack. Zero-trust. Zero-cost.
git clone https://github.com/VibeCodingLabs/openclaw-deploy.git
cd openclaw-deploy
bash install.shThe interactive CLI wizard will:
- Detect your system — OS, RAM, GPU (NVIDIA/AMD/CPU-only), storage, Docker status
- Install prerequisites — Node.js, Docker, Docker Compose, OpenClaw, uv, git
- Onboard OpenClaw — Walk you through
SOUL.md,AGENTS.md,USER.md,IDENTITY.mdcreation - Deploy your stack — Open-WebUI, Ollama, ChromaDB, n8n, OpenObserve (memory-budgeted for your hardware)
- Set up the Agent workspace — Full directory structure with RAG pipeline, security policies, heartbeats, and cron jobs
- Run health checks — Verify every service is alive and properly configured
openclaw-deploy/
├── install.sh # Interactive CLI wizard (start here)
├── .soul/ # Soul Spec — agent identity layer
│ ├── SOUL.md # Who the agent IS (philosophy, personality, boundaries)
│ ├── IDENTITY.md # How the agent APPEARS (name, emoji, voice)
│ ├── AGENTS.md # How the agent OPERATES (rules, memory, security)
│ ├── USER.md # Who YOU are (preferences, context, timezone)
│ ├── MEMORY.md # What the agent REMEMBERS (long-term facts)
│ └── BOOTSTRAP.md # First-run setup instructions (self-destructs after)
│
├── Agent/ # Agent capability modules
│ ├── LLM/ # Model routing (API keys, OAuth, providers)
│ ├── Sub-Agents/ # Agent hierarchy (Skills, Orchestration, Templates)
│ ├── Tools/ # External tools (CLI, MCP servers, scripts)
│ ├── Computer-Use/ # Browser/desktop automation
│ ├── Files/ # File processing (structured + unstructured)
│ ├── Multi-Modal-Prompt/ # Prompt engineering (text, vision, audio, system)
│ ├── Memory/ # 4-layer memory (short-term, long-term, vector, cache)
│ ├── Security/ # Zero-trust enforcement layer
│ ├── RAG/ # Retrieval-Augmented Generation pipeline
│ └── AutoResearch/ # Karpathy-style autonomous improvement loops
│
├── Heartbeat/ # Service health monitoring
│ ├── heartbeat.sh # Push-model heartbeat daemon
│ ├── healthcheck.py # Pull-model health endpoint aggregator
│ └── heartbeat.yaml # Configuration (intervals, thresholds, alerts)
│
├── Cron/ # Scheduled tasks
│ ├── Jobs/ # Active cron job definitions
│ ├── Logs/ # Cron execution logs
│ └── Templates/ # Reusable cron templates
│
├── Docker/ # Container configs
│ ├── docker-compose.yaml # Production stack (auto-tuned for your hardware)
│ └── .env.example # Template for secrets
│
└── Scripts/ # Utility scripts
├── setup-prereqs.sh # Prerequisite installer
├── onboard-openclaw.sh # OpenClaw onboarding wizard
├── setup-workspace.sh # Workspace structure generator
├── health-check.sh # Full stack health verification
└── teardown.sh # Clean uninstall
| Pain Point | How We Fix It |
|---|---|
| "Install takes 2 hours of debugging" | One wizard handles everything with rollback on failure |
| "SOUL.md? AGENTS.md? Which file does what?" | Guided onboarding explains each file and generates templates |
| "Docker OOMs my laptop" | Auto-detects RAM and sets memory limits per service |
| "Services die silently" | Heartbeat daemon with circuit breaker and auto-restart |
| "No idea if my stack is healthy" | health-check.sh tests every service endpoint |
| "Secrets in plaintext everywhere" | All secrets via env vars, fail-closed, never hardcoded |
| "RAG pipeline is basic" | v2 pipeline: hybrid search, reranking, semantic chunking |
| "Can't improve my agent prompts" | AutoResearch loops optimize skills while you sleep |
- Linux (Ubuntu/Pop!_OS/Debian) or macOS
- 8GB+ RAM (16GB recommended)
- Docker and Docker Compose
- Internet connection (for first-time setup)
MIT