Claude Code VS Code extension vs the CLI

We make Shuriken. We tried to be fair; corrections welcome at office@shuriken.sh.

Quick answer#

The Claude Code VS Code extension and the CLI are the same agent from Anthropic with two faces. The extension adds an editor interface: in Manual mode, Claude's edits as side-by-side diffs you accept or reject; your selection as context, plans you can review, conversations in tabs. The CLI keeps every command and skill, plus scripting and headless runs. Same account, same limits, same history. Most people use both.

Neither is ours. Both are made by Anthropic, so this is a page about picking the right face for the job, not about a winner.

At a glance#

Every cell is from Anthropic's VS Code docs, unless marked, read on 22 October 2026.

What the extension does well#

• Diffs where you read code. In Manual mode, when Claude wants to edit a file, the extension "shows a side-by-side comparison of the original and proposed changes, then asks for permission". You can edit the proposal before accepting, and Claude is told you changed it. From v2.1.275 you can accept or reject each change on its own. In Auto or Edit automatically mode, Claude edits most files without asking, so pick Manual from the mode indicator under the prompt box if you want every diff.
• Selection is context. Highlight code and Claude sees it; the prompt footer shows how many lines. Option+K (Mac) or Alt+K turns it into an @-mention like @app.ts#5-10.
• Plans you can edit. Review and change Claude's plan before it starts.
• Tabs. "Open in New Tab" or "Open in New Window" starts another conversation, each with its own history and context.
• Checkpoints with a mouse. Hover a message to rewind code to that point, fork the conversation from there, or both.
• MCP without the terminal. /mcp in the chat panel adds, removes and signs in to servers.

If you live in VS Code and want to see every change before it lands, the extension is the comfortable choice.

What the CLI does well#

• Everything. Every command and skill. The extension's docs say it plainly: "Some features are only available in the CLI."
• Scripts and pipelines. claude -p "prompt" runs without the interactive prompt, for CI, pre-commit hooks or a loop over files, with --output-format json or stream-json for parsing. The extension has no equivalent. Our Claude Code headless mode guide covers it.
• Anywhere there's a shell. A server over SSH, a container, tmux on a machine you leave running. No editor needed.
• Small speed-ups. ! to run a shell command in place, Tab completion, ↑ for history.
• Any editor. Zed, Vim, JetBrains or none at all. From an outside terminal, /ide connects it to a running VS Code for diffs and selection.

If you script things, work on remote machines or don't use VS Code, the CLI is the one. The extension can reach a remote machine too, through VS Code's Remote-SSH, WSL or dev container windows, but it installs and signs in on that side: Claude Code with VS Code Remote-SSH, WSL and containers.

Where they differ in practice#

The extension isn't an install of the CLI#

This trips people up. The extension bundles a private copy of the CLI for its panel, and "installing the extension does not put claude on your shell PATH". If you want claude in a terminal, run the standalone install too. Once both are in, claude mcp add and claude --resume work from any terminal, and both write the same settings.

Running the CLI inside VS Code gets you most of both#

With the extension installed, open VS Code's terminal (Ctrl+` or Cmd+`) and run claude. Anthropic's docs: "The CLI automatically integrates with your IDE for features like diff viewing and diagnostic sharing." You keep every command and still get diffs in the editor and your selection as context. The price is the terminal interface instead of the panel.

Moving between them#

Start a conversation in the panel, then run claude --resume in a terminal and pick it up there. "The extension and CLI share the same conversation history." Settings in ~/.claude/settings.json (allowed commands, hooks, MCP servers) apply to both.

VS Code extension or CLI: which should you use?#

• You live in VS Code and want to see each change before it lands: the extension, in Manual mode.
• You want several conversations side by side in the editor: the extension, in tabs.
• You script Claude, run it in CI or on a server: the CLI.
• You need a command the panel doesn't list: the CLI, in VS Code's own terminal.
• You use another editor, or none: the CLI.
• You can't decide: install both. They share everything, so trying one costs nothing.

Pricing#

Both are included with the same Claude plans, and neither costs extra. As of 22 October 2026, Anthropic's pricing page lists Claude Code on Pro ($20 a month, $17 billed yearly) and Max (from $100 a month), in US dollars; the free plan doesn't include it. The extension also works with a Claude Console account, billed at API rates.

The limits are shared. Anthropic's help page: "IDE usage counts toward the same usage limits shared across Claude and Claude Code." Switching from the panel to the terminal never gives you more room. Our beginner's guide explains how those limits work.

Using them together#

The common setup is both: the extension for focused work in the editor, the CLI for scripts, servers and the commands the panel lacks. They share an account, settings and history, so there's nothing to keep in step. Signing out in the panel signs the terminal out too, and a different account per project takes a VS Code profile with its own CLAUDE_CONFIG_DIR.

Shuriken sits next to both rather than replacing either. It runs Claude Code on your own computers, with each new chat on its own branch in its own git worktree. In the desktop app, Open in › VS Code opens a chat's worktree, or a file at its spot, in VS Code, where the extension is waiting; Cursor, Zed and the JetBrains IDEs are on the same menu, listed in Open anywhere. If you're coming from Cursor instead, see switching from Cursor to Claude Code.

To open your Claude Code chats in VS Code with one click, download Shuriken.