skip to content

Surfaces & Remote Access

The CLI, Desktop, VS Code and JetBrains, cloud sessions on web and mobile, and Remote Control or teleport between them. Asked because local vs cloud execution changes what the agent can reach.

part ofClaude Coderead it →
on this pageshow

explore

questions

6

Which Claude Code capabilities exist only in the CLI, and what do Desktop and the VS Code and JetBrains integrations offer instead?

level: juniorimportance: must knowfreq 26%

answer

  1. same engine, different front ends
  2. scripting and the Agent SDK
  3. Desktop is interactive only
  4. the extension's private CLI copy
  5. a plugin that launches claude

basics

~20 s

Non-interactive claude -p runs and the Agent SDK are CLI-only, and the CLI has every command and permission mode. Desktop adds visual diffs, panes and parallel sessions, VS Code a graphical panel, and the JetBrains plugin runs the CLI inside the IDE.

solid answer

~50 s

All of them run the same engine and read the same settings files, `CLAUDE.md` and MCP configuration, so a rule or memory file applies everywhere locally. The **CLI** is the most complete surface: `-p` scripting and the Agent SDK exist only there, it has every command and every permission mode including `dontAsk`, and it runs on Bedrock, Agent Platform and Foundry. **Desktop** trades that for visual review: a diff viewer, arrangeable panes, parallel sessions in a sidebar, Local, Cloud, SSH or WSL environments, and on Pro and Max computer use (research preview) and Dispatch; it is interactive only and has no agent teams. The **VS Code extension** is a graphical panel that bundles a private CLI copy, offers a subset of commands and no `!` shortcut, and does not put `claude` on your PATH. The **JetBrains plugin** bundles nothing: it launches your installed `claude` in the IDE terminal and adds IDE diffs and selection sharing.

go deeper

for a junior

Recall that every surface runs the same engine, that print mode and the Agent SDK are CLI-only, and that the VS Code extension does not install the claude command.

for a middle

Explain what each front end trades: Desktop's panes and environments against the CLI's full command and mode set, and why the JetBrains plugin needs the CLI installed.

for a senior

Standardise a team across surfaces through the shared settings files and CLAUDE.md, and know which gaps (providers, modes, agent teams) push a workflow back to the CLI.

for a principal

Decide which surfaces to support centrally: one shared configuration lowers drift, but provider support and scripting needs set a floor of CLI installs you must manage.

## One engine, several front ends Claude Code runs the same underlying agent on every surface: the **CLI** in a terminal, the **Code tab of the Claude Desktop app**, the **VS Code extension**, and the **JetBrains plugin**. The local surfaces also read the same configuration: `~/.claude/settings.json` and the project's `.claude/settings.json`, `CLAUDE.md` files, MCP servers from `~/.claude.json` and `.mcp.json`, hooks and skills. Concretely, these travel between local surfaces without any extra step: - **Settings files**: user, project and local `settings.json`, including permission rules and the `env` block. - **Memory**: the project's `CLAUDE.md` and `CLAUDE.local.md`. - **MCP servers** from `~/.claude.json` and `.mcp.json`; Desktop additionally reads `claude_desktop_config.json`, which the standalone CLI does not. A permission rule written once therefore applies whether the session was opened in a terminal, in Desktop or in an editor. What differs is the front end, and with it which features are reachable. ## What only the CLI does - **Non-interactive runs**: `claude -p "..."` (print mode) and the Agent SDK's scripting are CLI-only. Desktop states plainly that `--print` and `--output-format` are not available because it is interactive only. - **Every command**: other surfaces support a subset; commands that open a terminal panel, such as `/permissions`, have no argument form in Desktop and reply that they are not available there. - **Every permission mode**: the CLI offers all of them, including `dontAsk`; Desktop's selector offers Manual, Accept edits, Plan and Auto, plus Bypass permissions once it is enabled. - **Third-party providers**: the CLI runs on Amazon Bedrock, Google Cloud's Agent Platform and Microsoft Foundry; Desktop talks to Anthropic's API by default and needs a separate third-party deployment for those. - **Agent teams** (experimental) run in the CLI, not in Desktop. - **Computer use from the terminal** (a research preview on macOS for Pro and Max) is switched on through `/mcp` as the built-in `computer-use` server; Desktop offers its own version on macOS and Windows. ## Desktop: visual review and parallel sessions Desktop is the surface for people who would rather not live in a terminal. It adds a **diff viewer**, panes you drag into any layout (chat, diff, browser, terminal, file, plan, tasks), image and PDF attachments, sessions as **sidebar tabs**, and an **environment** choice per session: **Local**, **Cloud**, **SSH** to a machine you manage, or a **WSL** distribution on Windows. On Pro and Max it also offers **computer use** (a research preview on macOS and Windows) and sessions spawned by **Dispatch** from your phone. A CLI session moves into Desktop with `/desktop`. ## The IDE integrations | | VS Code extension | JetBrains plugin | | :- | :- | :- | | Interface | Native graphical chat panel; `useTerminal` switches to the terminal UI | The CLI running in the IDE's integrated terminal | | Ships its own CLI | Yes, a private copy for the panel | No; needs `claude` on your PATH | | Puts `claude` on PATH | No | Not applicable; you install the CLI first | | Notable extras | Inline diffs, plan review, @-mentions with line ranges, checkpoints | IDE diff viewer, selection sharing, diagnostics through `getDiagnostics` | | Gaps against the CLI | Subset of commands, no `!` Bash shortcut, no tab completion | It is the CLI in a terminal, so it has the CLI's feature set | Two consequences catch people out: 1. After installing only the VS Code extension, typing `claude` in a terminal fails: the bundled copy is private to the panel. Install the standalone CLI to use CLI-only features in VS Code's integrated terminal, where it integrates with the IDE for diffs and diagnostics automatically. 2. The JetBrains plugin shows "Cannot launch Claude Code" when `claude` is not on your PATH, because it does not bundle one. From a terminal outside the editor, `/ide` connects a CLI session to the running IDE. ## Choosing a surface - Scripting, CI glue or the Agent SDK: the **CLI**. - Reviewing many diffs, running several sessions side by side, or working over SSH from a window: **Desktop**. - Staying inside the editor with inline diffs: the **VS Code extension**; inside a JetBrains IDE, the **plugin** (which is the CLI). - Reaching any of them from a phone is a separate question, answered by Remote Control or a cloud session. Mixing surfaces on one project is normal: the shared settings and memory mean a team can standardise configuration once, while each developer picks the front end they work best in.

  • You installed the VS Code extension; why does typing `claude` in its integrated terminal fail?
    The extension bundles a private copy of the CLI for its chat panel and does not add `claude` to your PATH. Install the standalone CLI once; then `claude` works in any terminal and, inside VS Code's terminal, integrates with the editor for diff viewing and diagnostics. From an external terminal, run `/ide` to connect to VS Code.
  • Does a permission rule in `~/.claude/settings.json` apply to sessions in the Desktop app?
    Yes. Desktop and the CLI read the same settings files, so permission rules apply to Desktop sessions too, along with `CLAUDE.md` and MCP servers from `~/.claude.json` and `.mcp.json`. Desktop additionally loads servers from `claude_desktop_config.json` into local Code-tab sessions; the standalone CLI does not read that file.

  • Installing the VS Code extension also puts the claude command on your PATH.
  • Desktop can run non-interactive print-mode jobs just like the CLI.
  • The JetBrains plugin ships its own built-in copy of Claude Code.
  • Each surface keeps its own separate settings files and CLAUDE.md.
  • Agent teams work in the Desktop app exactly as they do in the CLI.
open as a page

In Claude Code, how does a Remote Control session differ from a cloud session in where it runs and what it can reach?

level: middleimportance: must knowfreq 22%

basics

~20 s

Remote Control lets claude.ai or the Claude app drive a session that keeps running on your machine, with your files, tools and MCP servers. A cloud session runs on a fresh clone in an Anthropic-managed VM and continues after you disconnect.

open as a page

How do `claude remote-control`, `claude --remote-control` and `/remote-control` differ in Claude Code, and what makes a Remote Control session go offline?

level: middleimportance: should knowfreq 9%

basics

~20 s

claude remote-control runs a server with no local prompt that can serve many sessions; claude --remote-control starts one interactive session also reachable remotely; /remote-control connects the session you are in. Each goes offline when the machine sleeps or the claude process stops.

open as a page

A Claude Code cloud session can't see your user-level MCP servers, skills or AWS SSO login — what does it actually inherit and reach?

level: seniorimportance: should knowfreq 11%

basics

~20 s

What the repository carries: CLAUDE.md, .claude/ skills, agents and rules, and in a one-repository session its settings.json and .mcp.json. User-level files, locally added MCP servers, plugins and interactive SSO never arrive, and outbound network defaults to the Trusted allowlist.

open as a page

From the Claude Code CLI, how do `--cloud`, `--teleport` and `/desktop` move work between surfaces, and what does each carry or leave behind?

level: seniorimportance: should knowfreq 9%

basics

~20 s

claude --cloud starts a new cloud session from your GitHub remote's current branch; --teleport pulls a cloud session's branch and conversation into the terminal; /desktop moves the CLI session into Desktop. No CLI command pushes a running session to the cloud.

open as a page

What does starting Claude Code with `--chrome` let the agent do, and what keeps the Chrome integration switched off?

level: middleimportance: nice to knowfreq 6%

basics

~20 s

claude --chrome connects the session to the Claude in Chrome extension, so Claude can open tabs, click, type, read console output and use sites you are signed into. It stays off on an API key or setup-token login, third-party providers and WSL.

open as a page