Switch model & mode — CodeAgent Mobile Wiki
The CodeAgent Mobile Wiki

Every feature, explained — organized by the path you actually take: set up, launch a session, supervise the run, ship the PR.

Free

Switch model & mode

Change the agent's model and its permission mode mid-session, right from your phone.

01 · MODEL
Pick the model
  • Whatever the agent advertises
  • e.g. Opus · Sonnet · Haiku
02 · MODE
Pick the mode
  • Default · Plan · Accept edits
  • Bypass permissions
03 · LIVE
Applied instantly
  • No restart, same session

You don't have to commit to one model or one permission level for a whole session. From the session controls you can swap the model the agent runs and its operating mode on the fly — the change is applied live to the running agent, no restart.

The in-session controls showing the mode selector "Default (recommended)" and a "Bypass Permissions" toggle
The in-session model + mode controls.
Switch the model

The app lists exactly the models your agent advertises — each with its context window — and switches to your pick live.

Switch the mode

Pick the agent's operating mode: Default (asks before acting), Plan (thinks without editing), Accept edits, or Bypass permissions — whatever it advertises.

Bypass permissions = fast, but trusted-only

Bypass permissions (Auto) auto-approves every tool call without asking. It's the fastest, but reserve it for workflows you fully trust.

Applied live

Both changes go straight to the running agent over the command relay — the same session keeps going on the new model or mode.

  1. 1
    Open the session controls

    The current model + mode are shown right in the session.

  2. 2
    Tap the model or mode

    A sheet lists what the agent supports, with the current one checked.

  3. 3
    Pick a new one

    The change is dispatched to your running agent and applied immediately.

TIP

The exact models and modes on offer are advertised by the agent itself — Claude Code exposes Default / Plan / Accept edits / Bypass permissions, and its available models — so the list always matches what your agent actually supports.

FAQ

Related

Ready to try it?