Has self-hosting open-source AI tools become harder than it should be?

I’ve been looking at how founders are using open-source AI tools lately.

The tools themselves are getting better every month. Agents, workflow builders, chat interfaces, internal assistants, automation layers — there is a lot you can do without starting from scratch.

But the part that still feels weirdly painful is everything around the tool.

Not the idea.
Not the use case.
Not even the model.

The setup.

You find a promising open-source AI tool, then suddenly you’re dealing with a VPS, Docker, SSL, environment variables, updates, backups, monitoring, storage, auth, and the fear that something will break when you actually start depending on it.

For technical founders, this is manageable but still annoying.

For non-technical founders, it can be enough to stop them completely.

I’m curious how others here think about this.

If you’ve tried self-hosting an AI agent, automation tool, or open-source AI app:

What was the most painful part?

Was it the first deployment, keeping it running, backups, updates, security, or something else?