Most coding agents today run on a laptop. An engineer installs Claude Code or Codex, opens a terminal, and gets a great experience for exactly one person.[1] Shut the lid and the session stops. The context the agent built over an afternoon lives on one machine, so handing the work to a teammate means starting over.
The bigger cost is that nobody else can see any of it. Leaders are approving real token budgets and asking a fair question: what are we getting for this? A team that says it is crushing it with AI might be getting a 2x speedup. That sounds fine, but if 10x is available you want 10x, and you can't close a gap you can't measure, which is the trap behind the 4x mandate. Centralized sessions make the question answerable. I ran a rough first pass on a week of my own team's sessions to see whether we were using agents for offense, building new features, or defense, protecting what we already shipped. I expected offense. We were doing far more defense. I suspect a lot of teams assume they're mostly on offense too. I would never have seen that from scattered terminals, and it sharpened my view on pointing agents at more than new features.
The cloud also changes what a session can be. Give every session the whole stack, already running, and the rice bowls go away: a backend engineer can fix a frontend bug because the frontend is right there, and no service depends on the one person who knows how to start it. Sessions become save points, like in a game. I can start something in the afternoon, log off, and a teammate picks it up that evening, switches to a different model, and keeps going with the full history. Publish the running app from each session and you get a staging environment per session, so a reviewer clicks through the product instead of reading a diff. It is the same instinct behind reviewing the session, not just the diff.
The usual objection is security, and I think it runs the other way. Isolated sessions with an audit trail are easier to govern than agents on hundreds of laptops sitting next to whatever credentials happen to be there, which is why I keep arguing to get the credentials off the laptop.
Design went through this already. Designers saved files locally, dropped them in Dropbox, and emailed versions around until Figma moved the work into a shared, live canvas.[2] I expect most coding, including at large companies, to happen in the cloud within a year, and plenty of startups already want all of it there. If your agents still run one laptop at a time, pick one real workflow this week, run it in a shared session with the whole stack up, and read the session log on Friday.
Key takeaways
- A coding agent on a laptop is a great experience for one person and invisible to everyone else.
- Shared cloud sessions let leaders see whether agent spend is buying a 2x or a 10x team.
- A session with the whole stack running can be paused, handed off, previewed and reviewed by anyone.
FAQ
Why move coding agents off the laptop?
Local sessions stop when the machine sleeps, are hard to hand off, and leave no shared record of what the agent did. Cloud sessions persist, can be picked up by a teammate, and can be measured.
Isn't running coding agents in the cloud less secure?
Usually it is the opposite. Isolated sessions with an audit trail are easier to control than agents running on hundreds of laptops next to whatever credentials are stored there.
Related Essays
Every CTO Is Asking How to Move Engineering Work to the Cloud
Startups are moving all of their engineering work into cloud agent environments. Enterprises are moving some. The constraint that decides how is the codebase itself.
Synchronous Cloud Development Is the Next Wave
Three months ago every startup wanted async background agents. Now the leading teams want all their coding sessions in the cloud, synchronous work included.
The IDE Is Going Away, and the Cloud Is Where Agents Run
Local machines cannot run ten agents at once. The cloud dev environment stops being a nice-to-have the moment agentic coding becomes the default.