Building on the go
Dispatch, remote control, and cloud agents: your laptop, from your phone.
The strangest realization once agents do the building: you don't have to be at the computer. A few surfaces make that literal, and they compound. This is where the “fixed subscription equals a 24-hour employee” thinking from our engineering meetups starts to apply to you.
Dispatch
Pair the Claude mobile app with your desktop (scan a QR code, two minutes) and you can assign tasks from your phone that your laptop executes, with its files, its connectors, its tools. Text your machine from the bus: “take the notes folder from today's calls and draft the weekly update.” It pings you when it's done. Same caveat as scheduled tasks: the laptop must be awake.
Remote control and cloud sessions
Claude Code sessions aren't chained to one screen. A build you start in the terminal can be checked and steered from your phone or any browser, and web-started sessions can be pulled down to your machine later. Codex goes furthest here with its cloud mode: tasks run on OpenAI's servers, laptop closed, three bugs fixed in parallel while you're at lunch. The pattern across all of them is the same: kick off long work, walk away, review when it's done.
Claude in your browser
The Claude for Chrome extension puts the agent inside your browser, where it can click, type, and read pages like a person. For a builder, the killer use is testing your own site: “open my deployed app, sign up as a new user, and tell me everything that felt broken or confusing.” It's also handy for the grunt work around a launch, like filling directory submission forms or pulling competitor pricing into a table.
Beginners use agents like a faster keyboard. Regulars use them like staff: work gets assigned, runs unattended, and comes back for review. Every surface in this chapter exists to make the second mode normal. The Agentic workflows section, near the end, takes it further.
AI Clubhouse · Vibe Coding Guide