OpenClaw — a personal multi-agent crew
A self-hosted crew of AI agents — a main assistant plus specialists for content, research, and health — running on my own infrastructure with a multi-provider fallback chain.
OpenClaw is a personal agent system I run on my own Ubuntu server. It is built around a main assistant that delegates to specialists: a content-drafting agent, a weekly research agent, and a health-coaching agent. Each has its own prompt, tools, and schedule.
The agents reach me through the channels I already use — Telegram and WhatsApp for chat, Gmail and Notion for work products — so there is no separate app to check. Everything is managed by systemd units, with health checks and restart policies, so the crew keeps running without me babysitting it.
The interesting part has been the operational side: handling provider rate limits, building a model fallback chain so a single provider outage does not take the crew down, and keeping auth tokens fresh across services. It is a small system, but running it has taught me a lot about what “in production” actually means for agents.