Setting Up My Personal AI Assistant with Clawdbot

January 25, 2026

I've been fascinated by the idea of having a personal AI assistant that I can talk to anytime, anywhere — not just through a web interface, but through WhatsApp, voice messages, or any other channel I use daily. Today, I finally set it up, and I want to share my experience.

Why Clawdbot?

There are many AI assistants out there, but most of them live behind a web UI or require you to open an app. What I wanted was something different: an AI that feels like a real assistant — one I can message on WhatsApp at 2am with a voice note, and it just... works.

Clawdbot is exactly that. It's a gateway that connects Claude (my AI of choice) to various messaging platforms. But more importantly, it gives the AI agency — it can read files, run commands, access my GitHub repos, create calendar events, and much more.

The Setup Journey

Getting Clawdbot running on my VPS was surprisingly smooth, though I learned a few things along the way.

The basics were easy: SSH into the server, run clawdbot doctor to check the state, and start the gateway. Where it got interesting was making it accessible from anywhere while keeping it secure.

I didn't want to expose my AI assistant to the public internet with just a password. Instead, I used Tailscale to create a private network between my devices. The clever trick? I pointed my custom domain to the Tailscale IP instead of the public IP. This means only devices on my personal VPN can reach it — everyone else gets nothing.

For SSL, I used Caddy which handles Let's Encrypt certificates automatically. No more messing with certbot renewals or nginx configs. Just point and shoot.

What Blew My Mind

Once everything was connected, I started testing features and honestly got a bit carried away.

Voice messages just work. I can send a WhatsApp voice note, and Clawdbot transcribes it using Groq's Whisper API (which is insanely fast — we're talking sub-second transcription). The AI then responds like it's a normal conversation.

It can write to my blog. I asked it to create a test post, and within seconds, it committed a new markdown file to my GitHub repo. This very post you're reading? Drafted entirely through a WhatsApp conversation.

Linear integration. I gave it my Linear API key, and now it can see my tasks, create issues, and help me stay organized. No more context-switching between apps.

Lessons Learned

A few things I'd do differently or want to highlight:

  1. DNS propagation takes time — I spent 20 minutes debugging SSL issues before realizing my DNS changes just hadn't propagated yet. Patience is a virtue.

  2. Tailscale for private access is genius — Pointing your domain to a Tailscale IP instead of a public IP is such an elegant solution. You get custom domain + SSL + private access without complex authentication.

  3. Caddy simplifies everything — If you're still manually managing SSL certificates, stop. Caddy handles it automatically.

  4. Persistent services matter — Creating proper systemd units ensures everything survives reboots. I learned this the hard way after my gateway died and I didn't notice for hours.

What's Next?

This is just the beginning. Here's what I'm planning to set up:

The goal is to build something that feels less like a chatbot and more like a digital extension of myself — one that can handle the tedious stuff while I focus on what matters.

Conclusion

Setting up Clawdbot took me about a day, but the result feels like a glimpse into the future of personal computing. Having an AI that's always available, understands context, and can actually do things in the real world changes how I think about productivity.

If you're technical enough to run a VPS and curious about AI assistants, I'd highly recommend giving it a try. The setup docs are solid, and the community on Discord is helpful.

Now if you'll excuse me, I need to go do some yoga. My AI has been reminding me for the past hour. 🧘

This post was drafted collaboratively with Clawdbot via WhatsApp.