Are new Start9 service a solution for AI agents - request for clarification of future products

[degard]

@matt

I just listened to your interview with BTCSessions regarding “
“No one knows what’s about to happen”.

I’m one of those who plan to run AI agents on a dedicated privacy machine.

I myself already use Start9 OS for my BIP110 node, mempool, etc. on a separate machine (NitroPC).

I intend to get a second machine as soon as possible to install the following setup (example):

Claude Code CLI + Claude Code Router + Ollama Server + qwen 2.5 (or something similar)

I’m NOT a true IT professional, so I find what Start9 is planning extremely exciting!

Unfortunately, I didn’t quite understand how my example setup could be configured with your new software solutions.

Therefore, I discussed this with ChatGBT as follows:

“ChatGPT: StartOS is primarily a server operating system for services from its marketplace, not ‘Windows/Linux for everyday use’ on which you can simply install any programs like your setup.

My recommendation is:

Bitcoin Server (1): Leave it as it is: just the Bitcoin stack and the components that absolutely need to be stable 24/7.

Second computer (2): Normal Linux (or Windows) as the main system and StartOS in a closed environment solely for server services, if you want maximum freedom for your coding setup (Start9 mentions virtualization as an option, but that adds more effort).”

:@Matt: Would you be so kind as to answer the following questions for me:

a) Is StartOS 0.4 capable of running a setup like Claude Code CLI + Claude Code Router + Ollama Server + qwen 2.5 (or something similar)?

b) Wouldn’t it be risky to run this setup on the same computer that hosts the Bitcoin services (1)?

c) Where should the planned router be installed? Do you provide a device for this?

d) What is your assessment of Chatgbt’s comment?

e) When exactly can we expect StartOS 0.4 and the router to be available? (Since I’m “very late” anyway, I’m planning to purchase a second computer in March/April 2026 and run AI agents on it.)

f) Would one of your hardware computers (server one, pure) be suitable as a second AI computer? (my current StartOS runs on a Nitro PC)

I have great confidence in Start9 and completely agree with your views on Bitcoin and privacy, and would therefore very much like to use your products (if it makes sense).

I look forward to your reply (if possible by refering to the numbers a-f).

Thanks!

Cheers, Mark

[StuPleb]

Hey @degard

I’ll see if I can answer some of this for you…

Claude Code CLI + Claude Code Router + Ollama Server + qwen 2.5 (or something similar)

Unfortunately, I didn’t quite understand how my example setup could be configured with your new software solutions.

In building StartOS v040, we’re laying the foundations for anyone/everyone to package services for StartOS, which could of course include AI agents packaging software for StartOS. The important (practical) part to note here is that services must be packaged in a way StartOS can understand and are integrated into the system. This includes any AI setup that itself could package software. What this means is that this example setup you mention, where you have Claude Code CLI running on StartOS, with Claude Code Router intercepting and sending everything to Ollama running Qwen… is all entirely possible but needs to be set up by someone initially. I don’t quite have a vision of the architecture of this… I suppose CLI+Router could be a single package that could point to Ollama, which already exists on v040… but the point is, it’s coming. Let’s see exactly what form it takes.

What ChatGPT said I imagine is related to some other extended conversation? A general “it’s good to separate a production machine running bitcoin from other stuff” is broadly correct.

@degard

a) Is StartOS 0.4 capable of running a setup like Claude Code CLI + Claude Code Router + Ollama Server + qwen 2.5 (or something similar)?

Totally. Particularly due to the fact that we’re working on GPU support so that locally run models can be sped up somewhat. Nothing comparable to a billion dollar datacenter… but much better than what is possible now.

@degard

b) Wouldn’t it be risky to run this setup on the same computer that hosts the Bitcoin services (1)?

In as much as it might be risky to do welding next to piles of paper cash.

You might be really careful, but you should probably fine a separate workspace.

What you mention here isn’t the riskiest setup though. It’s just Claude Code… you might be mixing this up with the wildly reckless openclaw/claudbot, which is basically a “whatever goes” experiment where agents have root access to everything via the same APIs, rather than a neat command line tool limited to specific directories.

@degard

c) Where should the planned router be installed? Do you provide a device for this?

The planned router that we will sell is a physical device. It will run an open source OS that you can flash on compatible devices.

@degard

d) What is your assessment of Chatgbt’s comment?

Makes sense to separate experiments and working space from important services, funds etc.

@degard

f) Would one of your hardware computers (server one, pure) be suitable as a second AI computer? (my current StartOS runs on a Nitro PC)

Well, we’re adding GPU support for running local models… so… you might be able to make a guess as to what we’d be selling eventually.

I’m no Matt, but I hope I’ve answer these well enough.

[degard]

Hi StuPleb, thanks a lot!

”What this means is that this example setup you mention, where you have Claude Code CLI running on StartOS, with Claude Code Router intercepting and sending everything to Ollama running Qwen… is all entirely possible but needs to be set up by someone initially.”

Honestly, I have no clue how this works. But I already managed to set it up on my local windows notebook with the help of ChatGBT.

I am now nearly ready to order a Meerkat (Meer10) with Intel Core Ultra 7 255H that Claude Sonnet 4.6 recommends and will be happy to then install StartOS 0.4, and later get a router as well.