Claude Code Beyond the Terminal — Web, Mobile, Desktop
The same engine runs in the CLI, the desktop app, the browser, on your phone and inside your IDE. What each surface can do, how work moves between them, and which plan unlocks what.
Chapter 8 | Claude Code Beyond the Terminal Written for: people already using Claude Code in a terminal who want work to continue when they are not sitting at the machine
The previous edition of this guide treated Claude Code as a terminal program and nothing else. That is no longer the whole picture. The same engine runs in several places, and each is tuned for a different way of working.
Where it runs
| Surface | Where the work actually happens | Best for |
|---|---|---|
| CLI in a terminal | Your machine | Everything; the most complete feature set |
| Desktop app | Your machine | Parallel sessions, visual diffs, a browser and editor pane |
| Web at claude.ai/code | A cloud machine | Work that must continue after you close the browser; needs the repo on GitHub |
| Mobile (Claude app, Code tab) | Nowhere — it is a client | Starting, watching and steering work while away from your desk |
| VS Code extension | Your machine | Inline diffs, @-mentioning files |
| JetBrains plugin | Your machine | The IDE’s own diff viewer and diagnostics |
| GitHub Actions / GitLab CI | A CI runner | Automated PR review, trigger-driven work |
What mobile actually is — the common misunderstanding
The mobile app is not where code runs. The documentation says so directly: the Claude app for iOS and Android “is a client for Claude Code sessions rather than a place where code runs”.
From your phone you can reach three things, and they differ in where the work runs:
| Route | You connect to | Use when |
|---|---|---|
| Cloud session | A cloud machine, Anthropic-managed by default | The repo is on GitHub and the task should keep going after you put the phone away |
| Remote Control | A session running on your own computer | The work needs your local files, tools or MCP servers |
| Dispatch | The desktop app on your computer | You want to message a task and let Dispatch run it. Requires a Pro or Max plan |
The rule of thumb is simple: if the computer will be off, use a cloud session. The other two drive your own machine, which has to stay on. If your machine sleeps mid-way through a Remote Control session, Claude Code reconnects when it wakes.
Type /mobile in a session for a download QR code (/ios and /android do the same thing).
The limitations worth knowing
- Commands that only run in the terminal interface, such as
/pluginand/resume, do not work from the app - The permission modes on offer are incomplete: cloud sessions get Accept edits, Plan and Auto; Remote Control gets Manual, Accept edits and Plan. You cannot select Bypass permissions from the app in either case, and you cannot select Auto for a Remote Control session
- Dispatch needs Pro or Max, and is not available on Team or Enterprise
The web — cloud sessions
Open claude.ai/code, pick a repository and branch, describe the task. Each session runs in its own isolated virtual machine, and it continues even if you close the browser. Pick it up later from your phone or back at your desk.
You can start one from the terminal too:
claude --cloud "a description of the task"The trap people hit: the cloud machine clones from the GitHub remote at your current branch, not from the files on your disk. Commits you have not pushed are invisible to it. Push first.
Bringing work back to your machine — teleport
A task started in the cloud can be pulled into your terminal to continue:
claude --teleport # pick from a list
claude --teleport <session-id>Or type /teleport (short form /tp) inside the cloud session itself.
Four conditions must hold before teleport will work: your working directory must be clean with nothing uncommitted, you must be in a checkout of the same repository, the cloud session’s branch must have been pushed to the remote, and you must be signed in to the same claude.ai account.
The desktop app
What it adds over the terminal: parallel sessions with git isolation, draggable panes (chat, diff, browser, terminal, file editor, plan, tasks, subagent), visual diff review, and pull-request monitoring.
The desktop Browser pane uses a clean browser profile, separate from your personal browser, with none of your saved logins or history. It suits testing the app you are building, not sites that need your identity.
macOS ships a universal build, Windows has x64 and ARM64 installers, and Linux is in beta.
IDE extensions
VS Code needs version 1.94.0 or later, installs from the marketplace, and works in forks such as Cursor through Open VSX. The point that confuses people: the extension bundles its own CLI for the chat panel, but running claude in the integrated terminal still needs the standalone CLI installed separately.
JetBrains works with IntelliJ IDEA, PyCharm, Android Studio, WebStorm, PhpStorm and GoLand. The plugin does not bundle a CLI — it runs claude in the IDE’s terminal, so install the CLI first. If you use Remote Development, install the plugin on the remote host, not on your local machine.
What each plan includes
This table comes straight from the official feature-availability page, and applies to people signed in with a claude.ai account. If you authenticate with an API key or through Bedrock, Vertex or Foundry, skip it.
| Feature | Pro | Max | Team | Enterprise |
|---|---|---|---|---|
| Claude Code on the web | ✓ | ✓ | ✓ | ✓ (needs a premium or Chat + Claude Code seat) |
| Routines | ✓ | ✓ | ✓ | ✓ |
| Remote Control | ✓ | ✓ | Admin-enabled | Admin-enabled |
| Channels | ✓ | ✓ | Admin-enabled | Admin-enabled |
| Computer use | ✓ | ✓ | ✗ | ✗ |
| Dispatch (desktop) | ✓ | ✓ | ✗ | ✗ |
| Code Review | ✗ | ✗ | ✓ | ✓ |
| Artifacts | ✓ | ✓ | ✓ | Admin-enabled |
| Analytics dashboard | ✗ | ✗ | ✓ | ✓ |
| Server-managed settings | ✗ | ✗ | ✓ | ✓ |
| SSO | ✗ | ✗ | ✓ | ✓ |
| SCIM | ✗ | ✗ | ✗ | ✓ |
| Zero Data Retention | ✗ | ✗ | ✗ | ✓ (enabled separately) |
What works on every model provider
If you are not on a claude.ai account but on the Anthropic Console, Amazon Bedrock, Google Cloud’s Agent Platform or Microsoft Foundry, everything that runs locally still works: the CLI, the Agent SDK, the VS Code and JetBrains extensions, subagents, hooks, commands, skills, CLAUDE.md, plugins, MCP servers, checkpoints, sandboxing and workflows.
What requires a claude.ai account is Claude Code on the web, on mobile and in Slack, the desktop app, Routines, Remote Control, the Chrome extension, computer use, Artifacts and voice dictation.
The documented workarounds when you are on another provider: if
/schedule(Routines) is unavailable, use/loopinstead; if cloud sessions are unavailable, use GitHub Actions or GitLab CI/CD.
What this chapter settles
Claude Code is no longer a terminal program alone, but each surface is limited in a different way, and the real limits come from three things: your plan, how you authenticate, and the surface itself. Choose by asking one question — where does this work need to run? If it needs local files, it needs your machine. If your machine will be off, it needs the cloud.
Glossary for this chapter
| Term | Meaning |
|---|---|
| Cloud session | A session running on a cloud virtual machine rather than yours |
| Remote Control | Driving a session running on your machine from another device |
| Dispatch | Sending a task from mobile for the desktop app to run |
| Teleport | Pulling a cloud session back into your terminal to continue |
| Admin-enabled | A feature that stays off until an organisation admin turns it on |