I love Clawdbot 🦞

Hello everybody, welcome back to Moodboard 🌴

You may have seen Clawdbot (now rebranded to Moltbot) blowing up this past week. It's been all over X and LinkedIn.

I'll be honest. I have, to this point, been largely skeptical of the available suite of agentic AI applications. 

Agents that book flights and order pizza? I mean, yeah, cool demos, but who actually needs (or wants to use) AI for that? 

The hype around autonomous AI always felt buzzy, buggy, and disconnected from how I actually work. The ‘agents’ I did try, I was fairly disappointed by.

So, when Clawdbot started trending, I was a little skeptical.

This weekend, I tried it (a lot), and it fully changed my mind 👇

What Is Clawdbot?

First, let's back up: what even is an "AI agent"?

When you use ChatGPT or Claude, you're having a somewhat siloed conversation. You type, it responds, you copy-paste the output somewhere. 

It forgets everything the moment you close the chat. You're initiating and orchestrating the work, the AI is just helping you execute each specific step you assign to it.

An AI agent is different. Instead of just chatting, it thinks, plans, and proactively takes actions.

It can run tasks on a schedule, access your files, fetch live data from the web, send messages to your team. 

It remembers who you are, what you're working on, and what you've asked it before. 

It feels less like a chatbot and more like a junior team member with specific permissions.

Clawdbot is an open-source framework for building exactly that - a personal AI agent. 

It can run locally on your machine or on a server, and you choose the brain: Claude, OpenAI, or any other model (I'm running Opus 4.5, which is a bit more expensive but gives it unreal performance + capability). 

Clawdbot connects to your tools (Slack, Telegram, local files, calendar, the web, etc.) so that it’s always accessible and feels fully equipped to handle any task you throw at it.

What makes it different from other agentic applications I’ve messed around with? It’s really just the simple stuff:

• It’s intuitive: you talk to it naturally, it figures out what to do (and, with your permission, does it)

• It’s functional: it actually works. It doesn’t feel like a tech demo. Smart usage results in serious time saved.

• It’s customizable: Clawdbot adapts to your workflows, your tools, your style. It remembers your preferences and compounds its memory over time.

• It’s universal: not a single-purpose agent for one niche task. It handles pretty much whatever you throw at it.

Use Cases

Using Clawdbot doesn’t feel like prompting, it feels like delegating.

After hours of experimentation (and ~$150 or so in token spend 🥴), I’ve found that, for the work that my team does, Clawdbot excels at:

1. Content Creation & Curation

It researches topics, pulls relevant sources, and drafts content in my format. Not generating from scratch, curating and assembling from real information it finds in the course of its research.

I stay in the loop for editorial decisions. It handles the legwork.

2. Orchestrating & Scheduling Workflows

I set up tasks that run on a schedule. It wakes up, does the work, and delivers the output to me.

There’s no need to remember to start the process (which can occupy more mental space than one might initially think!).

I review and approve rather than initiate and execute.

3. Team Delegation via Slack

It sends messages, updates, and deliverables directly to my team. I'm not the bottleneck copying and pasting outputs.

Clawdbot integrates into Slack (where my team already works), and keeps me in the loop while reducing my manual touchpoints.

I Get the Hype

I’m fully aware that I’m only scratching the surface of what Clawdbot is capable of, but after 3-4 days of intentional experimentation, I’m sold on a future where agentic applications like this are commonplace.

It really is something that you have to try for yourself. The feeling of your agent just “getting it” when you ask it to perform a multi-step task across multiple apps/platforms is really something else.

Clawdbot is like having a high-performing AI virtual assistant, one that works 24/7 and is powered by the most capable AI models.

I’ve already delegated hours of work to my agent, and I’ve been hugely impressed by the output.

Now, it's not plug-and-play yet. There's some setup involved, and you need to be comfortable working a bit in the terminal.

There are also important security concerns that you should take into consideration before granting an AI agent access to your computer + accounts.

You need to be smart and thoughtful with the permissions that you grant it, and you should at least be aware of the risks associated and safest practices for configuration.

But if you're comfortable tinkering, Clawdbot is open source and free to try: docs.clawd.bot

(And here’s a great YouTube tutorial for getting Clawdbot set up on a VPS)

If you're not ready to dive in, keep an eye on this space. OpenAI, Anthropic, and Google are all building their own agent frameworks. More polished, accessible options are on the horizon!