Plan Mode
Think first, code second. Plan Mode makes Claude research your codebase and write a plan you approve before a single file changes. It's the cheapest bug fix there is.
What is Plan Mode?
Plan Mode Plan Mode A read-only permission mode in Claude Code: Claude can explore the codebase and write up a plan, but can't edit files or run changes until you approve. Enter it by pressing Shift+Tab until the status bar shows plan mode, prefixing a prompt with `/plan`, or starting with `claude --permission-mode plan`. "Like walking through a house with an architect before renovation. You discuss ideas and make plans without swinging a hammer." Claude Code Anthropic's agentic coding tool. It lives in your terminal (and in VS Code, JetBrains, and on the web), reads your codebase, edits files, runs commands, and ships code. Install with the native installer (`curl -fsSL https://claude.ai/install.sh | bash` on macOS/Linux, `irm https://claude.ai/install.ps1 | iex` on Windows); it needs a Pro, Max, Team, Enterprise, or Console account. "Like having a senior developer living in your terminal, ready to help 24/7."
Think of it like a contractor walking through your house with a notepad before anyone picks up a hammer. You can still say "no, not that wall".
Why it matters: a wrong assumption costs one sentence to fix in a plan, and an afternoon to unwind in code.
Where Plan Fits Among the Permission ModesPermission Modes
Settings that control how much Claude Code can do without asking: `default` (shown as Manual, asks before edits and commands), `acceptEdits` (auto-approves file edits), `plan` (read-only planning), `auto` (Claude judges what's safe; the starting mode on Pro, Max, and Team plans), `dontAsk`, and `bypassPermissions` (no prompts at all; for sandboxes only). Shift+Tab cycles through the common ones.
"Like parental controls for AI. You choose how much freedom to give based on the task."
Permission Modes
Settings that control how much Claude Code can do without asking: `default` (shown as Manual, asks before edits and commands), `acceptEdits` (auto-approves file edits), `plan` (read-only planning), `auto` (Claude judges what's safe; the starting mode on Pro, Max, and Team plans), `dontAsk`, and `bypassPermissions` (no prompts at all; for sandboxes only). Shift+Tab cycles through the common ones.
"Like parental controls for AI. You choose how much freedom to give based on the task."
| Mode | Claude can... |
|---|---|
| auto | Proceed on its own for most actions. Pro, Max, and Team sessions start here. |
| default ("Manual") | Read freely, ask before edits and commands |
| acceptEdits | Edit files without asking; other actions still ask |
| plan | Read and research only, then present a plan |
The Shift+Tab cycle depends on where you start. From auto, the first press goes to Manual, then acceptEdits, then plan, then back to Manual. (Bypass and auto rejoin the cycle after plan when they're enabled.) So don't count presses: press until the status bar shows ⏸ plan mode on.
Three Ways In
1. Prefix one prompt with /plan
The quickest option. Plan just this request.
In a Claude Code session
/plan add Stripe subscriptions with a customer portal and webhook handling2. Toggle mid-session
Press Shift+Tab until you see ⏸ plan mode on.
3. Start the whole session in plan mode
Terminal
claude --permission-mode planBonus: scripted analysis
Combine with headless mode Headless Mode Running Claude Code non-interactively with `claude -p "prompt"`. Add `--allowedTools` to pre-approve tools, `--output-format json` (or `stream-json`) for machine-readable output, and `--bare` for CI. Perfect for scripts, GitHub Actions, and other automation. "Like leaving a note for your assistant instead of having a conversation. They do the task and leave the results."
Terminal
claude -p --permission-mode plan "Analyze how auth works in this repo and list risks" > auth-review.mdWhen to Use Plan Mode
Use It For
- •Multi-file changes (touching 3+ files)
- •Anything with auth, payments, or database schema
- •Complex refactoring
- •Learning a new codebase
- •Architectural decisions
- •Security reviews
Skip It For
- •Typo fixes
- •Single-line changes
- •Simple, well-understood bug fixes
- •Copy and styling tweaks
The Workflow
Explore
Claude reads your code
Plan
Proposes an approach
Refine
You question and push back
Execute
Approve, then it builds
Leaving plan mode: when the plan is ready, Claude asks whether to go ahead. Approve it to start building, or say no and keep refining. You can also press Shift+Tab to switch modes yourself.
Refining well: ask "what are you assuming?", "which files will change?", "how will we test this?", and "what's the simplest version?".
Getting Better Plans
Use the opusplan alias
Opus does the planning, Sonnet does the building. Strong reasoning where it counts, faster execution after.
/model opusplan
Raise the effort
Planning is where deep thinking pays off. Turn it up for the plan, down for routine execution.
/effort high
Hand it the spec
Point Claude at the requirements, the relevant files, and your constraints. Plans are only as good as their inputs.
Demand a test strategy
"Include how we'll verify each step" turns a plan into a checklist Claude can prove it finished.
Handy Shortcuts
| Shortcut | Action |
|---|---|
Shift+Tab | Cycle permission modes |
/plan <prompt> | Plan just this one request |
Ctrl+O | Toggle the transcript viewer (full tool calls, timestamps, model). It does not turn on thinking. |
Option+T / Alt+T | Toggle thinking on models that support it. No effect on Opus 5.5 or Fable, which always think. |
/effort | Set reasoning depth: low, medium, high, xhigh, max, or auto |
Esc Esc | Rewind to an earlier point in the conversation |
/resume | Continue an earlier planning session |
/compact | Summarize the conversation to free up context |
On older write-ups you'll see "Ctrl+O enables extended thinking Extended Thinking Letting the model reason step by step before it answers, which helps with architecture decisions and tricky bugs. Newer Claude models use adaptive thinking, where the model chooses how much to think (Opus 5.5 and Fable always think), while Haiku 4.5 still takes a manual thinking budget. In Claude Code, toggle thinking with Option+T (macOS) or Alt+T (Windows/Linux) and control depth with `/effort`. Ctrl+O opens the transcript viewer; it doesn't turn on thinking. "Like asking someone to 'think it through' before answering. They take longer but give better answers."
Plans love good context
A plan is only as good as what Claude knows going in. Learn to feed it well, then pick up the rest of the power-user habits.