[FEATURE]: Enabling dynamically changing agents and permissions while running prompt

[jmarcelomb]

Feature hasn't been suggested before.

• [x] I have verified this feature I'm about to request hasn't been suggested before.

Describe the enhancement you want to request

Hello, I like to be able to move between plan, build and autoaccept (yolo). However, I noticed that for example if I prompt with autoaccept I can click tab and the agent changes but it is not applied, only when doing a new prompt.

Is this already possible to change the agent permissions while working on a prompt?

My current config is:

{
  "$schema": "https://opencode.ai/config.json",
  "autoupdate": true,
  "permission": {
    "edit": "ask",
  },
  "lsp": {
    "ty-python": {
      "command": [
        "ty",
        "server"
      ],
      "extensions": [
        ".py",
        ".pyi"
      ]
    }
  },
  "agent": {
    "autoaccept": {
      "mode": "primary",
      "tools": {
        "write": true,
        "edit": true,
        "bash": true
      },
      "permission": {
        "edit": "allow"
      }
    }
  }
}

Thank you all! Marcelo

[github-actions[bot]]

This issue might be a duplicate of existing issues. Please check:

• #1813: Feature Request: Yolo and non-Yolo mode (auto-accept) - Requests the ability to switch between different permission modes (ask vs allow) similar to Claude's SHIFT+TAB cycling through Plan/Build/Yolo modes
• #5891: [FEATURE]: Full automation support - Requests full automation without manual confirmation for commands like rm, mv, copy
• #3261: Feature Request: Streamlined Agent Permission Management with an Interactive TUI - Proposes a TUI for managing agent permissions on-the-fly without editing config files

Feel free to ignore if none of these address your specific case.

[rekram1-node]

You have to submit a message to change the agent, otherwise it's just a toggle. You can even just send "." and itll change the agent mid loop

[jmarcelomb]

Would be that bad to add ability to when tabbing it could switch the agent dynamically? Or for example, after enabling for all, would be nice to have a /reset command, something like that

[jmarcelomb]

In the last update I like that we have a second prompt, no more writing auto accept mistakes, but we have This will allow edit until OpenCode is restarted.

Could it be possible to reset this in any point of time and not needing to restart? It would be very nice. Thank you for all the work!