From the sofa into the session
How I work with my agents at home from my phone – and which tools keep my everyday life running in the background.
Sometimes, while I am out or in bed, I think of something I want to change in a project. I do not want to have to go to my desk, find the right computer and work out which session is still open. I want to get the thought out while it is still clear.
My phone or tablet then gives me access to my computers at home. The actual work happens there.
I am less interested in a ranking than in the question: which small detour does each tool save me?
1. Termius: the terminal in your hand
Termius is my starting point on my phone or tablet. It gives me a terminal for connecting to another computer. Instead of opening my laptop for a quick task, I open the app and enter my working environment. I like having the same access to my sessions even when I am not sitting at my desk. The small display does not replace a large screen when reviewing an interface. For a task or a follow-up question, it is often enough.
2. tmux: the session stays at home
tmux keeps the terminal session open on the remote computer. I first connect via SSH and then attach to the tmux session there. This means I do not have to start again if my phone loses its connection or I close the app. I used to think of a conversation like this as tied to the open terminal window. Now I can leave it and pick it up again later. I like this separation: the connection can be short-lived, while the work stays in the same place. tmux does not protect against a computer restart or a crashed process.
3. Tailscale: the private way home
Tailscale SSH is the connection to the hub, my central computer at home. In this workflow, it replaces a publicly accessible SSH connection. I do not have to expose terminal access to the open internet. I like that the route goes through my private tailnet and that I can limit access permissions there. This access also needs suitable rules; “private” does not mean every device should automatically be allowed to do everything.
4. Start the Claude session in HQ
On the hub, I go into the HQ folder and start a Claude session there. I say which project and which server to work on. The agent reads the relevant profiles and confirms its scope. I have described the structure in the article How do I organise multiple AI agents across different servers?. This saves me searching for the right starting point and explaining which role each machine currently has. I like that even a new conversation history gets a fixed starting point.
Access is available only within the tailnet, not publicly; the agents run at home, not on the phone. tmux keeps the session open if the connection drops. On top of that come the limits of the HQ setup: this does not give a development agent unrestricted write access to production.
5. Wispr Flow: say the thought out loud
With Wispr Flow, I dictate instead of typing longer tasks on a small keyboard. I often know quite clearly what is bothering me, but would reduce it to a few bare keywords when typing. Speaking lets me include the context: what I am seeing, what should happen and why. That is what I like about it. I still read the text before sending it. A misrecognised word should not turn into the wrong task. Afterwards, I check follow-up questions and results just as I would at my desk.
What runs at home
Not every tool in my everyday life is part of this connection. Some run in the background and make sure I can find things again or have fewer things to operate individually. For me, they still belong in the same collection: software is useful when I do not have to keep thinking about it.
Paperless is my document store. Instead of looking for a file in lots of folders, I want to be able to find its contents and check details about it. I describe my own interface and the support for reading documents on the Paperless project page. I like the combination of an existing document store and a view that fits my everyday life. AI can help with sorting; I still need to be able to see whether a suggestion is correct.
Home Assistant connects devices and routines in my home. It removes some of the switching between individual manufacturers’ apps. On the Home Assistant project page, I show why I build my own additions for it. I like being able to describe a process from my point of view instead of starting again with each device. Not everything has to happen automatically. Sometimes a good shared overview is already an improvement.
UniFi is my tool for the home network. Instead of checking in different places, I want to see in one place whether the connection is basically working. This is the unremarkable foundation that makes the route from the sofa work at all. I like an overview that helps when something goes wrong. The specific cabling, devices and access details, however, do not belong on this page for me.