Claude Code plan mode: how it works, and when to use it

Plan mode vs auto mode vs accept edits vs manual

Plan mode is one of Claude Code's permission modes. Each trades convenience for oversight; the table shows what Claude can do without asking you in each.

How to use plan mode in Claude Code

1. Enter plan mode

   In the terminal, press Shift+Tab to cycle modes until the status bar shows ⏸ plan mode on; from Manual the cycle is accept edits, then plan. Or start a single prompt with /plan. In VS Code, click the mode indicator under the prompt box and pick Plan; in the desktop app, use the mode selector.

   # Start a session in plan mode
   claude --permission-mode plan
   
   # Or, inside a session, plan just this prompt
   /plan add rate limiting to the public API

2. Describe the change and let Claude explore

   Ask for what you want built or fixed. Claude reads the code and runs commands to investigate, then writes a plan. When auto mode is available, a classifier reviews those commands instead of asking you; otherwise anything beyond the built-in read-only commands asks first.

3. Review and edit the plan

   Answer Claude's questions, push back, or press Ctrl+G to open the plan in your text editor and change it yourself. Plans are saved as files in ~/.claude/plans/, or the folder set by the plansDirectory setting.

4. Approve, or keep planning

   Choose Yes, and use auto mode (or Yes, auto-accept edits when auto mode isn't available) to let Claude implement it; Yes, manually approve edits to review each edit; or No, keep planning. Approving leaves plan mode and names the session after the plan.

5. Leave without approving

   Press Shift+Tab again to exit plan mode without a plan. To plan the next change, cycle back or start the prompt with /plan.

Source: Claude Code docs, Choose a permission mode

Make plan mode the default

For one session, start with claude --permission-mode plan. For every terminal session in a project, set defaultMode to plan in .claude/settings.json; for every project on your machine, put it in ~/.claude/settings.json.

1. Add it to settings

   VS Code conversations don't read project settings for the starting mode: set claudeCode.initialPermissionMode to plan in your VS Code user settings instead.

   {
     "permissions": {
       "defaultMode": "plan"
     }
   }

When plan mode is worth it

• Changes that touch many files, or where the first idea is often wrong: a migration, a refactor, a new feature across front end and back end.
• Code you don't know yet: let Claude map it and explain the approach before anything changes.
• Expensive mistakes: database schemas, auth, payments, public APIs.
• Skip it for one-line fixes and tasks you'd describe the same way as the plan; manual or accept edits mode is faster there.
• Pair it with a clean context: when showClearContextOnPlanAccept is on, approving can also clear the planning conversation so implementation starts from the plan alone.

Gotchas

• Plan mode blocks edits, not everything: with auto mode available, approved shell commands run during planning, so a command with side effects can still execute.
• In an interactive terminal session started with bypass permissions available, plan mode's blocks aren't enforced; Claude is only instructed not to edit.
• Headless runs (claude -p), Agent SDK sessions and the VS Code chat panel keep plan mode's blocks.
• Picking Plan in VS Code applies to that conversation only; new conversations go back to your usual starting mode.
• Deny rules still apply in every mode, plan included. See Claude Code slash commands for /plan, /permissions and other commands.

Plan mode keeps Claude careful. Crewmark keeps it working.

Plan mode is how you stay in charge of a big change; the session still waits for you to start the next one. Crewmark turns a Claude Code, Codex, DeepSeek Harness or OpenCode session into an AI employee with a role, a goal and working hours, and wakes it on a heartbeat to take the next issue from a board.

1. Plans as issue comments

   An employee can post its plan on the issue and wait for your go-ahead before a risky change, so you approve from the board instead of the terminal.

2. Let the heartbeat start the work

   During its working hours the employee wakes every 15, 30 or 60 minutes in the same conversation and takes the next issue; see how to make Claude Code proactive.

3. Several employees, one repository

   Give each its own worktree and role. Read what an AI employee is.

Claude Code plan mode FAQ