# Claude Code plan mode: how to use it, the shortcut, and how it compares with the other modes

Plan mode is where you get a say before any code changes. How to turn it on, what the approval choices do, and when Anthropic says to skip it.

Published 2026-10-07 by Richard Kaminsky and Mitchell Lipyansky, the co-founders of Poly. Canonical: https://usepoly.co/claude-code-plan-mode
Poly is a multiplayer AI coding workspace: a shared room where your team works with one AI agent, together. Free to start: https://usepoly.co/

**Plan mode makes Claude Code research and propose changes without making them: it reads files and runs commands to explore, then writes a plan and asks you to approve it before editing anything. Turn it on by pressing Shift+Tab until the status bar shows plan mode, typing /plan (optionally with the task, as in /plan fix the auth bug), or starting with claude --permission-mode plan. When the plan arrives you can approve it and let Claude work in auto mode, approve it but review each edit, press Ctrl+G to edit the plan yourself, or choose "No, keep planning" and say what to change.**

## Key takeaways

- **Read-only:** in Anthropic's words, plan mode "tells Claude to research and propose changes without making them."
- **Three ways in:** Shift+Tab, `/plan`, or `--permission-mode plan`; set `defaultMode` to `plan` to start every session there.
- **Approval:** "Yes, and use auto mode," "Yes, manually approve edits," or "No, keep planning." Ctrl+G opens the plan in your editor.
- **When to use it:** for changes that touch several files or code you don't know. "If you could describe the diff in one sentence, skip the plan."
- **opusplan:** a model setting that plans with Opus and builds with Sonnet.

*Claude Code's permission modes, from Anthropic's docs, October 7, 2026.*

| Mode | Edits files | Runs commands | Best for |
| --- | --- | --- | --- |
| Manual | After you approve | Most after you approve | Careful work in unfamiliar code |
| Accept edits | Yes | Simple file commands; others ask | Iterating on code you're reviewing |
| Plan | No, until you approve a plan | To explore | Multi-file changes and new code |
| Auto | Yes | Yes, with a safety classifier | Long tasks without prompts |
| Bypass permissions | Yes | Yes, no checks | Isolated containers only |

## What does plan mode do in Claude Code?

Claude explores your code (reading files and running commands to look around) but doesn't edit source files. It hands research to a built-in, read-only Plan subagent, then presents a plan for your approval ([Anthropic](https://code.claude.com/docs/en/permission-modes)). Where auto mode is available, a safety classifier approves exploratory commands instead of asking you each time.

Plans are saved as Markdown files in `~/.claude/plans` by default, named after your prompt; the `plansDirectory` setting moves them into your project, where a team can review them.

## How do you turn on plan mode?

- **Shift+Tab** cycles the modes: from auto it goes to Manual, then Accept edits, then Plan. On Windows, if Shift+Tab doesn't work, the key is Alt+M.
- **`/plan`** switches to plan mode from the prompt; `/plan fix the auth bug` starts planning that task at once ([Anthropic](https://code.claude.com/docs/en/commands)).
- **`claude --permission-mode plan`** starts a session in plan mode; it works with `-p` too.
- **By default:** set `"permissions": {"defaultMode": "plan"}` in `.claude/settings.json`.
- **VS Code:** pick Plan from the mode indicator, or type `/plan`; the extension opens the plan as a Markdown document you can comment on line by line ([Anthropic](https://code.claude.com/docs/en/vs-code)).
- **Desktop app:** choose Plan in the mode selector next to send. It applies to the current session only.

Press Shift+Tab again to leave plan mode without approving anything.

## How do you approve or change a plan?

When Claude finishes, it shows the plan with these choices ([Anthropic](https://code.claude.com/docs/en/permission-modes)):

- **Yes, and use auto mode:** Claude carries out the plan with background safety checks instead of prompts. Where auto mode isn't available, this reads "Yes, auto-accept edits."
- **Yes, manually approve edits:** Claude works through the plan but asks before each edit.
- **No, keep planning:** type what you want changed, and Claude revises the plan.

To change it yourself, "Press Ctrl+G to open the proposed plan in your default text editor and edit it directly before Claude proceeds."

## Plan mode vs auto mode vs accept edits

All of Claude Code's permission modes, from Anthropic's docs:

- **Manual** (`default`): reads freely, asks before edits and most commands.
- **Accept edits** (`acceptEdits`): makes file edits and simple file commands without asking; other commands still ask. This is what "auto accept" and "edit automatically" mean.
- **Plan** (`plan`): reads and explores, edits nothing until you approve a plan.
- **Auto** (`auto`): does everything, with a classifier reviewing each action and blocking risky ones such as piping a download into a shell or force-pushing. The starting mode for interactive sessions since v2.1.283.
- **Bypass permissions** (`--dangerously-skip-permissions`): does everything with no checks; Anthropic says isolated containers and virtual machines only.

Auto mode pauses and goes back to asking after 3 blocked actions in a row or 20 in a session, and an organization can turn it off. The full guide to the modes is in [How to use Claude Code](/how-to-use-claude-code).

## When should you use plan mode?

Anthropic's best practices recommend four steps: explore, plan, implement, commit ([Anthropic](https://code.claude.com/docs/en/best-practices)). "Letting Claude jump straight to coding can produce code that solves the wrong problem." Plan when you're unsure of the approach, the change spans several files, or the code is new to you; skip it for small, obvious edits. More in [Claude Code best practices](/claude-code-best-practices).

**opusplan** is a model setting for this split: it "uses opus during plan mode, then switches to sonnet for execution," so the planning gets the stronger model and the typing gets the cheaper one ([Anthropic](https://code.claude.com/docs/en/model-config)). Choose it in `/model`.

## Plan mode on the web, your phone and in Codex

- **Cloud sessions** (claude.ai/code and the Claude app) offer Accept edits, Plan and Auto; Remote Control offers Manual, Accept edits and Plan ([Claude Code on your phone](/claude-code-remote-control)).
- **Plan here, build in the cloud:** Anthropic's docs suggest planning locally, committing the plan, then starting a cloud session to carry it out ([Claude Code on the web](/claude-code-on-the-web)).
- **Codex** has a plan mode too: `/plan` or Shift+Tab in the Codex CLI ([OpenAI](https://learn.chatgpt.com/docs/developer-commands)).

Ultraplan, Anthropic's cloud planning preview, was removed in August 2026; plan mode replaces it.

## Approving a plan as a team

In Claude Code the plan comes back to the one person who asked for it. In Poly (usepoly.co), a team works with one Claude Code or Codex agent in a shared browser room: the plan appears for everyone, any member can approve it or send it back with changes, and every change after that can wait for a member's yes. Free to start. [What is Poly?](/what-is-poly)

## Common questions

**What is plan mode in Claude Code?**

A permission mode where Claude researches your code and proposes a plan without editing anything. You approve the plan, edit it, or ask for changes before Claude starts making edits.

**What is the shortcut for plan mode in Claude Code?**

Shift+Tab cycles the permission modes until the status bar shows plan mode; on Windows, Alt+M if Shift+Tab doesn't work. You can also type /plan or start with claude --permission-mode plan.

**What is the difference between plan mode and auto mode?**

Plan mode only reads and proposes; nothing changes until you approve. Auto mode makes changes and runs commands without asking, with a safety classifier blocking risky actions. A common pattern is to plan first, then approve the plan into auto mode.

**Can you edit a Claude Code plan before approving it?**

Yes. Press Ctrl+G to open the plan in your text editor and change it directly, or choose No, keep planning and tell Claude what to change. In VS Code you can comment on the plan line by line.

**What is opusplan in Claude Code?**

A model setting that uses Opus while in plan mode and switches to Sonnet to carry out the approved plan, so planning gets the stronger model and execution the cheaper one.

More guides: https://usepoly.co/guides · Security: https://usepoly.co/security · Pricing: https://usepoly.co/pricing