I've started playing with OpenClaw
First weeks with my feet on the ground: what it is, what I've achieved and what I haven't.
It depends on what you're after and how seriously you take permissions. OpenClaw is a free, open-source AI agent that you install on your own machine and…
It depends on what you're after and how seriously you take permissions. OpenClaw is a free, open-source AI agent that you install on your own machine and that, instead of just answering, does things: it reads and writes files, runs commands, browses websites or sends messages. It's worth it if you want to understand what an agent really is, but it isn't the most comfortable place to take your first steps with AI.
The noise around it doesn't help much. It was born in 2025 under another name (first Clawdbot, then Moltbot), took the name OpenClaw in January 2026 and within a few weeks became one of the most popular projects on GitHub, so there's no shortage of videos presenting it as the assistant that will run your whole life. What gets said less is that a program capable of acting on your computer can also make mistakes on your computer. Installing it without thinking about what you let it touch is the quickest way to get a fright.
I started trying it recently, with more curiosity than haste, and I have to fit it in alongside the agents I already use. What I've done so far, what went well and what didn't I describe in "I've started playing with OpenClaw", which is a first-impressions article that I'll keep updating. If you want to try it yourself, a practical tip: do it on a separate machine or a test VPS, never on the computer where you keep what matters.
First weeks with my feet on the ground: what it is, what I've achieved and what I haven't.
One email a week with what was verified, what was refuted and what is still on the bench. No news roundups.