What is OpenClaw?
A practical guide to the piece of software that massively changed how I work: what it is, why it matters, and how to get started.
In the last month I exchanged over 10,000 messages with an AI agent called Cove. It's built on something called OpenClaw and I interact with it through WhatsApp or Slack. He (_he?_) handles a growing chunk of my work -- email, calendar, expenses, meeting prep, research, code, life admin.
I didn't plan for it to become this central. I set it up out of curiosity and three weeks later it was just how I worked. I created a WhatsApp group for my birthday and the first message someone posted was "Was this group created using OpenClaw?"
I've used ChatGPT since 2022. I use Claude all day every day. They're great and I couldn't live without them. But they sit in a tab, waiting. This is different. It works alongside me whether I'm paying attention or not. That seems minor, but it's a really different experience and I believe it's the start of a huge shift in how we use and interact with AI.
I wrote this piece because I get asked the question "What is OpenClaw" on a daily basis. This is my attempt to get it down in one place (so that I can just send a link!). Here goes:
The software is called OpenClaw. Open-source, built by Peter Steinberger. You install it on a server and connect it to WhatsApp, Telegram, Discord, Slack, Teams, etc -- wherever you already message people.
There's no app to open. No tab to switch to. It's just there, in your messages, alongside your conversations with real people. The friction of "going to the AI" disappears entirely.
It has files -- a personality file, a memory file, notes about your preferences, your projects, your working style. Every conversation adds context. It's memory is better than something like ChatGPT, but there's room for improvement. I don't think we're far from the world where it knows how you think after a few days and anticipates what you need after a few weeks.
The thing that breaks people's mental model is that it acts on its own. Check my email every morning and flag anything urgent. Prep me before meetings. Track my expenses. Send me a summary of my week every Sunday. You tell it once. It just does it. In the background. While you sleep.
I run a company called Verascient with my co-founder Emile. Two people. I think we're doing way more than a two-person company should be able to do. But it doesn't feel like two people. Cove handles things that would normally require a part-time assistant, a bookkeeper, and an ops person.
Here's a normal morning. I wake up and there's already a message waiting -- my calendar for the day, urgent emails, a thinking prompt based on what I've been working on. Meeting at 10? There's a prep doc by 9:30 with context on who I'm meeting, what we discussed last time, and what I might want to cover.
I didn't build any of this. I told Cove I wanted it and it set everything up. When it needed access to my email, it walked me through connecting it. When it needed a new capability, it wrote the code itself.
It just figures things out and gets better day-by-day.
Last month I told Cove I wanted to track every receipt and invoice automatically. It connected to my Gmail, wrote parsing rules for different vendors, started downloading PDFs, extracting amounts, uploading them to Google Drive in organised folders, and logging everything to a spreadsheet. When I send it a photo of a restaurant bill on WhatsApp, it reads the receipt, categorises it, uploads it, and logs it.
I didn't install a plugin. I didn't configure an integration. I described what I wanted via a voice note and it built the thing.
Cove has become the layer that much of my work runs through. Everything still happens -- the emails, the meetings, the code, the admin -- but it runs through Cove instead of through my head.
Since you made it this far, something must have interested you, so I'm going to take the opportunity to try sell this to you.
Should you try it?
If you do knowledge work and spend any meaningful part of your day on coordination, admin, or context-switching -- yes.
Connect it to WhatsApp and start talking. No special syntax. No commands. You just tell it what you want the way you'd explain it to a new team member.
The first day feels like a novelty. By day three you'll start moving real work over. By the end of the first week you won't want to go back.
I know how that sounds. I'd roll my eyes reading it too. But 10,000+ messages in ~20 days doesn't happen because something is a novelty.
We're building Verascient to make it easy and safe for individuals and businesses to start using these kinds of agents. For individuals, the setup takes ten minutes. For businesses -- enterprise-ready deployment, security, compliance, the things that stop a company from letting every employee run their own agent. If that's interesting, I'd love to hear from you.
The gap between people who have this and people who don't is widening every day, purely because most people don't know it exists yet.
Now you do.