You're viewing documentation for unreleased changes. View the latest release.

Read-only Mode

Read-only mode is a read-only mode for exploring a project and asking questions without changing anything. While it’s active, the assistant can read and search everything, but cannot modify files or directly run R/Python code.

The guarantee: no workspace or home-directory writes, and no direct R/Python code execution. Shell commands are limited to known read-only commands.

Direct execution means running code through the console or notebook execution tools. Allowed Positron help lookups can still import Python modules and execute their top-level code.

Entering Read-only Mode

Type /readonly or pick Read-only from the mode picker in the status bar. Only you can enter Read-only mode — the assistant cannot enter it on its own.

What Still Works

Read-only mode blocks changes, not information. The assistant can still:

  • Read and search everything — reading files, listing directories, and grep/glob search all keep working, including browsing outside the workspace (with the normal permission prompt)
  • Inspect your R/Python sessions — reading console history and viewing plots
  • Look things up — web search and fetching documentation
  • Run read-only Positron commands — in Positron, commands that Positron marks as only reading something or changing what’s on screen, such as reading the Viewer. Which commands are marked depends on your Positron version
  • Propose changes in chat — if you ask for a change, the assistant presents it as a suggestion, diff, or code block for you to apply yourself

What’s Blocked

  • File edits and writes — editing, creating, or deleting files, including notebooks
  • Code execution — no direct R/Python code execution
  • Mutating shell commands — anything beyond known read-only commands (e.g. git status, ls, cat) is refused
  • Other Positron commands — any Positron command that Positron doesn’t mark as read-only
  • MCP tools — all MCP tools are unavailable while Read-only mode is active

Blocked actions are refused outright rather than confirmed — there is no approval prompt that would let them through, and approval modes like /yolo don’t override the restriction.

Exiting Read-only Mode

Type /normal or use the mode picker. Only you can exit Read-only mode — the assistant cannot exit it on its own either. If it needs to make a change, it will suggest that you exit Read-only mode first.

When to Use Read-only Mode

Read-only mode is most useful for:

  • Understanding unfamiliar code — explore a codebase without assistant-authored file edits
  • Reviewing before changes — discuss an approach or inspect state before letting the assistant edit
  • Restricted environments — work alongside the assistant (it writes code, you run it) where writes or execution are off-limits

For designing an implementation approach with a written plan, use Plan mode instead.