What Claude Code on the Web Actually Is
Claude Code on the web is the browser version of Claude Code. It lives at claude.ai/code, and instead of running on your laptop, the task runs in a cloud session on Anthropic-managed infrastructure, or on your organization's own infrastructure if that's been set up. The practical difference from typing claude in a terminal or opening the desktop app's Code tab is where the work happens: not on the machine in front of you, but on a virtual machine you can close the lid on.
The detail that actually changes how you work is persistence. A cloud session keeps running after you close the browser tab or shut your laptop. You can open the Claude mobile app later, see what it did, leave a comment on the diff, or send it a follow-up message, all without your laptop being open or even in the room.
How It's Different From the Desktop App's Code Tab
The two aren't competing products, they're the same underlying session reached from different places. Here's what actually changes depending on where you start it.
Where a task actually runs
| Desktop app (Code tab) | Claude Code on the web | |
|---|---|---|
| Runs on | Your own machine | A cloud VM (Anthropic's infrastructure, or your org's self-hosted environment) |
| Needs installed | Yes, the Claude desktop app | No. Sign in at claude.ai/code in any browser |
| Survives closing the window | No, the session pauses with your machine | Yes, the session keeps running |
| Where you can check on it | Only that machine | claude.ai/code, the terminal, or the Claude mobile app |
| Best for | Active building with a diff view and browser pane open in front of you | Tasks you don't need to watch happen |
If you already use the desktop app, this isn't a separate tool to learn from scratch. The Code tab's Cloud environment option starts the same kind of cloud session, see our [desktop app guide](/blog/claude-code-desktop-app) for how that fits into the three-tab workflow we teach. The web version at claude.ai/code is the same underlying session, reached from a browser instead of the app.
How to Start One, From the Browser or Your Terminal
Getting a repository connected is the only real setup step, and there are two ways to do it. During web onboarding at claude.ai/code, you can authorize the Claude GitHub App, which is also what unlocks auto-fix on pull requests (more on that below). Or, if you already work from a terminal, run /web-setup inside Claude Code to sync your existing gh CLI login instead, no separate app to install.
- From claude.ai/code: sign in, connect GitHub, pick a repository, and describe the task.
- From your terminal: run claude --cloud "describe the task" to send it to the cloud while you keep working locally. Use /tasks to check progress without leaving the terminal.
- Bring it back: run claude --teleport (or /teleport inside a session) to pull a cloud session into your terminal and keep going locally, once its branch is pushed and your working directory is clean.
The Off-Desk Test: When I Actually Send a Task to the Cloud
I still build most of my day in the desktop app, with the diff view and browser pane open where I can watch each change land. Before I reach for a cloud session instead, I run what we call the Off-Desk Test inside the club, three short questions, and if I answer yes to at least two, it goes to the cloud.
- Am I about to be away from this machine? A commute, a client call, an overnight run. If yes, the cloud session keeps working while the desktop app would just sit idle.
- Is the work already fully pushed to a branch? Cloud sessions clone from your GitHub remote, not your uncommitted local state, so anything still sitting in your working directory needs to be committed and pushed first.
- Can I review the result from wherever I end up? The diff view and mobile app cover reading and commenting, but if the task needs a full local dev environment to actually verify, running it locally is the shorter path.
The Off-Desk Test isn't about which tool is better. It's about matching where a task runs to whether you're actually going to be sitting in front of it. Most client work still belongs in the desktop app, in Manual or Accept Edits mode, with you watching. The cloud is for the parts of the day you're not at your desk anyway.
What You Get in a Cloud Session
- A diff indicator on the session, something like +42 -18, that opens into a full diff view, the same review habit as any local session.
- Auto-fix on pull requests, if you've installed the Claude GitHub App: Claude watches CI checks and review comments on a PR and pushes a fix when it's confident, or asks you when a comment is ambiguous.
- Sharing controls. Team and Enterprise accounts can set a session to Private or Team-visible; Pro and Max accounts can set Private or Public, and Public means any signed-in claude.ai user can open it. Check a session for credentials or client-sensitive code before you flip that toggle.
- Sessions expire after a period of inactivity and the VM gets reclaimed, but reopening it from claude.ai/code spins up a fresh VM with your conversation history restored, so nothing you did is actually lost.
The Self-Hosted Option, If Your Org Needs It
One more layer worth knowing exists, even if you'll rarely touch it as a solo builder: organizations on Team or Enterprise plans can route these same cloud sessions onto their own infrastructure instead of Anthropic's, through a public beta called self-hosted environments. It's off by default, and turning it on is a real infrastructure commitment, not a settings toggle. If a client or a bigger agency you work with asks about it, that's a pricing and scoping conversation of its own, we walk through exactly how to price that request in [how to price a Claude Code build when the client wants it self-hosted](/blog/pricing-self-hosted-claude-code-for-clients).
Try It on One Task This Week
Pick one task you'd normally start right before leaving your desk, something you'd otherwise just let sit until you're back. Push your branch, run claude --cloud with the task described in one clear sentence, and check it from your phone before you sit back down.
Short, practical drops on skills, MCP, agents, prompts, and more. No spam, unsubscribe anytime.
Frequently asked questions
What is Claude Code on the web?
It's the browser version of Claude Code at claude.ai/code. Instead of running on your own machine, the task runs as a cloud session on Anthropic-managed infrastructure (or your organization's own infrastructure, if self-hosted environments are set up). The session keeps running after you close the browser, and you can check on it, review the diff, or send a follow-up from the Claude mobile app.
Do I need the desktop app to use Claude Code on the web?
No. You can sign in at claude.ai/code in any browser and connect a GitHub repository without installing anything. If you already use the desktop app or terminal, the same cloud sessions are also reachable from there, including the claude --cloud command.
Does closing my laptop stop a cloud session?
No, that's the actual difference from a local session. A cloud session keeps running after you close the browser tab or shut your laptop. You can check its progress later from claude.ai/code, your terminal, or the Claude mobile app.
Is Claude Code on the web the same as the self-hosted option?
No, they're different layers. Claude Code on the web is the browser entry point at claude.ai/code, and by default those sessions run on Anthropic's own infrastructure. Self-hosted environments is a separate, opt-in beta for Team and Enterprise organizations that routes those same cloud sessions onto infrastructure the organization runs itself.
Should I use the cloud version instead of the desktop app for client work?
Not by default. The desktop app still gives you the full diff view, browser pane, and local review in one window, which is where most client work should happen. Reach for a cloud session when you genuinely won't be at your machine to watch it, not as a replacement for the review habits you'd use locally.
Last reviewed by David Iya on August 23, 2026
