VibeKit
Enter App
Claude Code Remote Control — Drive a Coding Agent From Your Phone
"Remote control" sounds complicated; it isn't. A coding agent never really runs on your phone — it runs on a computer somewhere, and your phone is the remote. The only decision that matters is where that computer is and who keeps it awake. There are two clean answers, and VibeKit does both.
What "remote control" actually means
When people say "control Claude Code from my phone," what they want is: kick off a task, walk away, glance at progress, steer it, and get the result — without sitting at a laptop. The agent loop (read files, edit, run commands, repeat) has to run on a real machine with a file system. So the setup is always the same shape: a Claude Code session running on a host + a mobile client driving it. The choice is which host.
Route 1 — Pair your own machine
Point your phone at a Claude Code session running on a computer you own. Anthropic's own Claude Code remote control works this way, and so does VibeKit's vibekit-agent CLI:
npm install -g vibekit-agent
vibekit-agent link # pair this machine with the iOS app
vibekit-agent start # listen for tasks against this directoryNow a task sent from your phone runs Claude Code against your real local codebase, commits, and reports back. Best when the work lives on your machine — the bug you noticed while away from the desk. The catch: the machine has to stay awake. Close the lid or lose power and the session is gone.
Route 2 — Run the agent fully hosted
The other answer is to not keep any machine alive at all. With a hosted agent, each project gets its own Claude Code–style agent that lives on a server 24/7, plus a git repo and a live *.vibekit.bot URL. You drive it from a native iOS app (or web, or CLI). Start a task, lock your phone, come back to it finished.
- Nothing on your end to keep awake — the agent runs server-side, always on
- Survives a dropped connection: subway, elevator, spotty wifi only disconnect your view, not the agent
- Each app keeps its own context, repo, and environment — separate agents, not one shared session
- Ships a real, shareable app at a public URL — not just a session you babysit
Which should you use?
| Pair your own machine | Fully hosted (VibeKit) | |
|---|---|---|
| Where the agent runs | Your laptop / desktop | VibeKit's cloud (always on) |
| Keep a machine awake? | Yes — sleeps = session dies | No |
| Survives a dropped signal? | Fragile mid-turn | Yes — reconnect to where it got |
| Public URL per project? | No (local-only) | Yes (*.vibekit.bot) |
| Best for | Driving your existing local repo from your phone | Starting + shipping projects entirely from your phone |
| Set up | vibekit-agent | iOS app |
Why hosted holds up better on a phone
The hardest part of controlling an agent from a phone isn't sending the task — it's connectivity. Phones drop signal constantly, and a remote-control link to a session on your own laptop is exactly where that hurts: lose the link mid-turn and the in-flight work can vanish. A server-side agent decouples the two — the agent keeps running on the host no matter what your phone's radio is doing, and the app just re-attaches to the live session when you're back. You can fire off a long task on the subway, surface in daylight, and read what it did.
Bring your own key
VibeKit is BYOK: bring your own Anthropic or OpenAI key and the agent bills your key directly with no markup, or use pay-as-you-go credits. There's a free tier to try it first. See BYOK for how the key drives the actual coding agent (not just AI features inside your built app).
FAQ
What does it mean to control Claude Code remotely?
The coding agent doesn't run on your phone — it runs on a computer somewhere (your own machine, or a server) and your phone is the remote that sends tasks and reads the output. "Remote control" is just a Claude Code session running elsewhere, driven from a mobile client. The only real choice is where that session runs and who keeps it alive.
Do I need to keep my laptop awake?
Only if the agent runs on your laptop. If you pair your own machine (Anthropic's Claude Code remote control, or VibeKit's vibekit-agent), the session dies when the machine sleeps. If you run the agent hosted on VibeKit instead, it lives on a server 24/7 — nothing on your end to keep awake, and you can close the app mid-task and come back to it finished.
Does it keep working if my connection drops?
With a hosted agent, yes — because it runs server-side, a dropped signal only disconnects your view and you reconnect to wherever it got. Remote-controlling a session on your own laptop is more fragile: if the link drops mid-turn you can lose the in-flight work.
Is this the same as Claude Code's dispatch / remote control?
Same goal — drive a Claude Code session from your phone — different model of where it runs. Anthropic's remote control points your phone at a session on a machine you keep running. VibeKit gives you both: a vibekit-agent CLI to pair your own machine, or a fully hosted agent that runs in the cloud with a native iOS app, a git repo, and a live URL per project.
Can I use my own Anthropic or OpenAI key?
Yes. VibeKit is BYOK — bring your own key and the agent bills it directly with no markup, or use pay-as-you-go credits. There's also a free tier to try it.
Related
- VibeKit Remote Agent — pair your own laptop and drive its Claude Code from iPhone
- Claude on mobile — the native iOS app for coding agents
- Code on iPhone — build and ship real apps from your phone