The OpenClaw FOMO finally won me over

The FOMO of having my own AI assistant finally caught up to me, and I am giving it a shot with OpenClaw. I'm rarely an early adopter of new tools and hardware, and I always steer clear of jumping on hypes right away. I prefer to let more excited people test things first before I spend time and money on them. But, a month and a half after the boom (for AI development, might as well be years), I think it's time.

What am I going to use it for, though?

I like doing my own writing, my own research, developing my own methodologies and making sites for my projects, so I probably won't use it for that on its own. Sure, I use Claude Code for most all of the weight lifting now, but I do have to think about solutions for my work and do enjoy writing pipes in R when analyzing data. And I certainly don't want (or need) AI to manage my life, like schedules, tasks, shopping or when my dogs need grooming. I like autonomy, and it does not involve a chatbot doing stuff around my personal life.

On the other hand, summarization is a big thing for me, reading complex papers in search of useful information might be one thing. I might also use it to monitor some stuff, like news, public documents, politicians etc. and have it write daily reports about the stuff I want to see directly in an Obsidian vault of my choosing. That could be useful.

I am concerned with security and privacy, so I was thinking about a good way to try to test how useful this can be to me. I won't install it into my personal computer, there is no chance of that. I don't want to buy a new Mac Mini or whatever new hardware this might require before I understand the potential upside this might have for me – otherwise I would risk only having one extra $600 computer without much use.

So it was decided that I needed some virtual private server, the cheapest one.

One option would be to get a platform as a service like Heroku or Replit, which are convenient, but come at the cost of, mostly, control of what I can install (no root access), always-on server, data migration and, sure enough, price points. Another would be to install a virtual machine on my computer, using something like UTM, but I have never done that before and I am unsure if I like the idea.

So I settled on getting myself a virtual private server somewhere. In Nucleo we use Hetzner for all our servers, but that comes way too "naked", and I would have to spend too much time in configurations, installing dependencies.

There comes Digitalocean. I've used DO many times before, and it is a reliable place for testing and even scaling. It is somewhat cheap, although if you stack a bunch of resources pricing can go up fast, as it happens on most of those server companies like AWS (you blink and then you have two Lightsails, seven EC2s and three RDSs running).

Digitalocean has a 1-click app for OpenClaw that is an absolute beauty, with very little friction, after I got the US$16/month (assuming it can handle it). I had Clawsula (that's what I named it) up and running almost immediately using DO's own terminal window (for some reason I failed to configure OpenClaw using ssh in my own computer).

My OpenClaw stack so far is this:

• Model: Claude Haiku 4.5, from Anthropic ($3 per million tokens)
• Search: Brave API ($5/month in credits, plus $5 extra that they give you)
• Messaging: Telegram, although I want to try Discord
• Text: I want Clawsula to write stuff in an Obsidian vault I created just for it

I finished all my setups today. Lets see how that goes.