Full Stack Web Developer, Sinan Eldem
I stopped running coding agents on my laptop. Here's the setup I ended up building (feedback welcome)
For most of this year I've used Claude Code and Codex every day on client projects. The agents got good enough that the bottleneck stopped being the model and became everything around it:
- a 30-40 minute refactor dies when I close the laptop lid
- I keep alt-tabbing to a terminal to check if it's done, failed, or sitting on a question it asked 20 minutes ago
- every project needs its own PHP/Node version, a database and a URL I can actually open, before the agent can test anything
- letting an agent run unattended without a checkpoint to roll back to makes me nervous
So the agents moved to a server I own. What's there now:
- agent sessions run on the server in the background and survive me closing the laptop
- each project gets its own runtime version, database and an HTTPS preview URL, so the agent can run the tests and I can click through the result
- a checkpoint/backup before a long run, so "undo everything it did" is one click
- a push notification on my phone when a session finishes, fails, or is waiting for my answer. This one changed my day the most: I answer the agent's question from my phone and it keeps going.
It started as a pile of scripts and turned into a product called rdev (disclosure: I built it and it's paid, plans start at $12/month). You keep root on your own server; only the panel is hosted. iPhone app is out, Android is in testing.
Honestly, you can get a lot of this with tmux on any VPS plus a notification hook, and that's how I started. What I'd really like to hear from people here:
- Where do your agents run today, laptop, dev server, or a cloud sandbox?
- What's the thing that still makes you babysit them?
- What would stop you from running agents on a server you own?
Agentsclaude-codeself-hostedfeedback






Comments
0 comments