Run Claude Code remotely: every option, honestly mapped
Laptop closed, agent still working: there are now six real ways to get there. A map of all of them — and the queue problem none of them solve.
Why “remote” became the default ask
Agentic coding sessions outgrew the laptop. A real task runs twenty minutes to a couple of hours; laptops sleep, Wi-Fi drops, and you have places to be. So the question everyone asks after a week of Claude Code is the same: how do I keep this running — and steer it — when I'm not at my desk? As of mid-2026 there are six real answers. They solve different problems, and picking wrong wastes an afternoon, so here's the honest map.
Option 1 — Claude Code on the web
The zero-setup answer: Claude Code runs as cloud sessions at claude.ai/code, in Anthropic's sandbox, from any browser — including the one on your phone. Kick off a task, close the tab, check back later. This is the right default if your work lives in cloud repos and you don't need your own machine's environment. The limits are the flip side: it's Anthropic's box, not yours — your local toolchain, secrets, and long-lived daemons aren't there.
Option 2 — Remote Control and Channels (official)
If the session runs on your machine, Anthropic's Remote Control lets the Claude mobile app steer it. Channels goes further: bridge a running session into Telegram, Discord, or iMessage — messages in, permission prompts and progress out. Free, official, and good. This is the “control Claude Code from my phone” answer for a machine you already keep on.
Option 3 — the mobile command centers
Omnara (polished mobile command center, push notifications, fleet view) and Happy Coder (free, open-source, end-to-end encrypted) wrap your sessions in a real phone UI rather than a chat bridge. If you juggle several concurrent sessions and want an app, not a thread, they earn their place. Honest comparisons: vs Omnara, vs Happy Coder, vs Channels.
Option 4 — the VPS classic
A ~$5/month box, tmux, your Claude plan: the setup that started the whole trend, and still the most flexible. Your environment, your daemons, always on. SSH from a phone works (Termius, Blink) but is fiddly enough that most people pair the VPS with option 2 or 3 for steering. The full recipe — and what to actually do with the box — is in Claude Code on a VPS.
Option 5 — GitHub as the remote runtime
No machine at all: install the Claude GitHub App and @claude mentions turn issues into PRs inside GitHub Actions. You “run Claude Code remotely” without owning anything that runs it. The full setup and its honest limits: Claude Code + GitHub.
Picking
- Just want tasks to survive the laptop lid → web sessions (option 1). Zero setup.
- Own machine, occasional steering from the phone → Remote Control / Channels (option 2). Free and official.
- Several sessions, want a real mobile UI → Omnara or Happy (option 3).
- Your own environment, always on, maximum control → VPS (option 4), steered by 2 or 3.
- No infrastructure, repo on GitHub → the GitHub App (option 5).
The sixth option is a different question
Everything above connects you to your agent. It all still starts with you noticing a problem, scoping it, and typing it — remote access moved the keyboard, not the bottleneck. The sixth option feeds the agent without you in the input path: chatwithdev puts a chat widget on your site, relays conversations to your Telegram, and when a customer reports something real you tap ➕ Ticket and dispatch it — to GitHub or to a runner on your own box:
npx chatrelay-runner init --token crr_… # guided setup, verifies the connection npx chatrelay-runner # polls for dispatched tickets
Progress streams back to the same Telegram: started → PR opened → merged. You review from your phone, and one more tap tells the customer it's fixed. Remote control is how you drive the agent; this layer is what keeps it busy while you don't.
Getting started
Whichever remote option you pick, the queue works with all of them: start free, one script tag, and your website chats arrive in Telegram — ready to become dispatched fixes when you are.
FAQ
Can I run Claude Code on my phone itself?
Not locally — phones can't run the CLI. What works from a phone: cloud sessions in the browser at claude.ai/code, steering a session on your own machine via Remote Control or Channels, or a mobile command center like Omnara or Happy Coder in front of it.
Do I need a VPS to run Claude Code remotely?
No. Web sessions and the Claude GitHub App both run without any machine of yours. A VPS earns its $5 when you want your own environment, your own daemons, and an agent that's always on — see the VPS guide for the full recipe.
What's the best way to control Claude Code from Telegram?
For interactive steering of your own session, Anthropic's Channels is official and free. For the customer-facing layer — website chats relayed to Telegram, tickets, and dispatching fixes with progress streamed back — that's chatwithdev; the two compose in the same Telegram app.
Which option keeps the agent busiest?
None of the remote-access options add work — they move your keyboard. If the goal is an agent that works while you don't, the input has to come from somewhere else. Customer reports are the highest-signal source: real defects, fresh reproduction details, reported the moment they happen.