Clawdbot / Moltbot: The AI Agent Everyone's Talking About

An AI agent called Moltbot (formerly known as Clawdbot) took over my social media timeline.

All over Twitter, LinkedIn, and YouTube, you’ll find people calling it "what Siri was supposed to be," and rushing to buy Mac Minis on which to securely run it (which I find both hilarious and intriguing at the same time).

Some of the demos are genuinely impressive: negotiating and purchasing cars, calling restaurants when OpenTable fails, managing your entire schedule through WhatsApp.

It feels like the future we were promised.

There's something deep going on here. But there are also a lot of people getting excited about something they probably shouldn't use yet.

So today, I want to break down:

• What Moltbot actually is and what makes it different

• Why most people should wait before setting it up

• My take on whether it's worth your time right now

—@dharmesh

What Moltbot Actually is

In short, Moltbot is an open-source AI agent that runs locally on your Mac and operates 24/7 within messaging apps like Telegram or WhatsApp.

The technical setup is pretty interesting: it's not running in the cloud somewhere. It's on your machine, with access to your local files, your apps, your accounts. And you control it entirely through chat.

But what makes it different? Well, most AI tools make you come to them. You open ChatGPT. You open Claude. You switch contexts, describe what you need, and copy-paste results back to where you were working.

Moltbot comes to you.

More importantly, it actually executes. When you ask it to do something, it doesn't just give you instructions or generate text. It can:

• Send messages on your behalf

• Manage and edit local files

• Browse the web and extract information

• Run code and terminal commands

• Book things through websites

• Set up recurring automations

It also remembers context across sessions. Moltbot builds on what it already knows about you and your preferences.

What we're seeing is a glimpse of what happens when an AI agent has the same access to your computer that you do. That's powerful. And also, as we'll get into, kind of terrifying if you don't know what you're doing.

Why I Think Most People Should Wait

Firstly, I don't recommend running Moltbot on your primary machine (yet). The project is too new, and the potential risk is too high. If you're going to experiment, use a secondary device or virtual machine.

Second, setting up Moltbot does require technical ability. You need to run terminal commands, set up environment variables, debug cookie authentication, configure API keys, understand cron syntax for scheduling, and manage system permissions.

Even with Claude helping you through it, this is not a "download and go" experience. If you don't already know what those terms mean, it’s still possible to set up, but it will take some time and more importantly, some patience.

But the third and largest issue is security.

You're giving an AI agent access to your accounts. It can read your messages, send texts on your behalf, access your files, and execute code on your machine.

Here’s a good list of vulnerabilities and fixes if you do plan to tinker:

Long story short, you need to actually understand what you're authorizing. It's fairly easy to trigger things you don't want -- like having the agent respond to texts from anyone in your contacts when you only wanted it to respond to you.

This is still early. The capabilities are impressive, and I like that it's open source. But my guess is that the frontier model companies (OpenAI, Anthropic, Google) are going to offer something along these lines soon -- with better security models, easier setup and recognized brands. It’s easier to explain to your IT people that you’re using OpenAI Co-conspirator (I just made that name up) than MoltBot.

For now, Moltbot gives us a glimpse of where personal AI agents are headed. But I think most non-technical people should wait for someone else to package it properly due to risks.

My Personal Take

I think Moltbot is important for what it represents, not necessarily for what it is today.

The capabilities are real. The technology (mostly) works. And it's open source, which means developers can build on it and learn from it. For technical users, it’s an incredible playground.

But for the average person? I think this is the situation of technology in search of a use case.

So should you try it?

If you're technical, enjoy tinkering, and understand the security implications mentioned above -- then yes. Set aside a weekend and experiment. You'll learn a lot about where agents are headed and maybe build yourself a personal assistant.

If you just want an AI assistant that works with little-to-no set up? Probably wait a couple more months. The capability Moltbot demonstrates will show up in consumer products soon with better security protections.

The hype is real in the sense that this shows us the future. But the reality is that most people aren't ready for this level of complexity and risk.

That gap will close. Just not this week. But things move fast, so maybe next week. :^)

—Dharmesh (@dharmesh)