OpenClaw, the powerful fresh agent assistant, has a weakness for guacamole.
This is one of several things I discovered last week when I used a viral AI bot as my personal assistant.
OpenClaw, formerly known as both Clawdbot and Moltbot, has recently become the darling of Silicon Valley, charming AI enthusiasts and investors looking to either leverage the cutting-edge technology or profit from it. The highly proficient, web-savvy AI bot has even inspired its own exclusively (or mostly) AI social network.
As a WIRED’s writer AI Lab newsletter, I thought I should take the risk and try using OpenClaw myself. I instructed the bot to monitor incoming emails and other messages, search for fascinating research, order groceries, and even negotiate deals on my behalf.
For brave (or perhaps reckless) early adopters, OpenClaw seems like a legitimate look into the future. However, all wonder is accompanied by a tinge of terror as the AI agent sifts through emails and file systems, wields a credit card, and sometimes even turns on the human user (though in my case, this distraction was entirely my fault).
How I set it up
OpenClaw is designed to run on a home computer that is always on. I set up OpenClaw to run on a Linux PC to access the Anthropic Claude Opus model and talk to me via Telegram.
Installing OpenClaw is plain, but setting it up and keeping it running can be a headache. You need to provide your bot with an AI backend by generating an API key for Claude, GPT, or Gemini, which you paste into your bot’s configuration files. In order for OpenClaw to operate Telegram, I also had to do that first create fresh Telegram bot, then provide your OpenClaw bot credentials.
For OpenClaw to be truly useful, you need to combine it with other software tools. I created a Brave Browser Search API account to allow OpenClaw to search the Internet. I also configured it to be able to access Chrome via an extension. And, God facilitate me, I gave him access to the email, Slack and Discord servers.
Once this was all done, I could talk to OpenClaw from anywhere and tell it how to operate my computer. First, OpenClaw asked me some personal questions and let me choose his personality. (The options reflect the anarchic feel of the project; my bot, Molty, likes to call itself a “gremlin of chaos”). The resulting personality is very different from Siri or ChatGPT, and this is one of the secrets of OpenClaw’s runaway popularity.
Internet research
One of the first things I asked Molty to do was to send me daily fascinating research articles in the field of artificial intelligence and robotics from the arXiv platform, where researchers submit their work.
I previously spent several afternoons vibration coding websites (www.arxivslurper.com AND www.robotalert.xyz) to search the arXiv. It was amazing (if a little demoralizing) to see OpenClaw instantly automate all the required review and analysis work. The items she selected are so-so, but I think it could be much better with further instruction. This type of network searching and monitoring is certainly helpful, and I think I will operate OpenClaw for this often.
IT support
OpenClaw also has an uncanny, almost terrifying ability to fix technical issues on your computer.
This shouldn’t be surprising given that it’s designed to operate a boundary model that allows you to easily write and debug code and operate the command line. Still, it’s amazing when OpenClaw simply reconfigures its own settings to load a fresh AI model or debug a browser issue on the fly.