OpenClaw — развёртывание AI-агента в production

★ 6.9 · marketing

openclaw-agent-deployment is a Claude Code skill that takes an OpenClaw AI agent from zero to a production-ready deployment: defining the agent's identity and rules in SOUL.md, setting up the runtime via Docker or pm2, connecting to Telegram, Slack, or Discord, and configuring an LLM provider — Claude 3.5 Sonnet, GPT-4o, Ollama/Llama 3.3, or DeepSeek V3. The skill walks through eight steps: authoring SOUL.md, preparing the environment, populating .env with API keys, attaching messaging channels, hardening the gateway on port 18789 (localhost-only with SSH tunneling for remote access), scheduling periodic tasks via HEARTBEAT.md, and setting up pm2 logs with cron-based health checks. It targets VPS, home servers, and Raspberry Pi deployments (512 MB RAM minimum; 4 GB+ for local Ollama). A library of 177 ready-made SOUL.md templates is available in the awesome-openclaw-agents repository.