I've been running OpenClaw for about three weeks now and I can't emphasise enough how much it has changed the way I work. This isn't a marginal improvement, it's closer to 100x for building things and easily 10x in terms of effectively having a small team just doing stuff for you in the background. It runs my blog operations, handles outreach for EabhaSeq, monitors my analytics, builds me dashboards, and has become the operational centre for everything I do day to day. The sky genuinely feels like the limit with this thing, and it's only just getting started.

For those unfamiliar, OpenClaw is an open source AI assistant created by Peter Steinberger that runs on your own hardware and connects through messaging apps like Telegram, iMessage and WhatsApp. It uses LLMs to execute real tasks through a skills and MCP server system, and critically it runs 24/7 with persistent memory and context. OpenAI just acquired it for a reported $1 billion after only 84 days of existence, which should tell you something about how significant this project is. I've found it to be the first AI tool that actually feels like having a team member rather than talking to a chatbot.

My Setup

Most people running OpenClaw seem to be on Mac Minis, and there's a good reason for that given the native macOS integration and browser automation. I went a different route and I'm glad I did. I run two instances: one on a small DigitalOcean droplet and another on my Linux desktop. The DO droplet handles the always-on tasks like email monitoring, scheduled research and analytics checks, while the desktop instance handles heavier interactive work when I'm at my machine. Total cost for the always-on instance is about $6 a month, which for what it does is absurd.

The entire interface is through Telegram. I message OpenClaw like I would message a colleague, and it messages me back with updates, questions and completed work. I'm using Claude as the LLM provider, and I've got somewhere between 10 and 15 skills installed: Gmail, Google Calendar, Plausible Analytics, OpenAI Whisper transcription, Brave search, browser integration, Telegram of course, and a custom dashboard skill that I'll get to shortly.