You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I have verified this feature I'm about to request hasn't been suggested before.
Describe the enhancement you want to request
When we plan something and it finishes, we have to switch to the build agent and then type in something to actually let it build. I find this unintuitive. Claude code for example gives the option to continue planning or start to build with a simple multiple choice. This I feel should be the way. I doubt anyone is switching to any other agent bar build after planning.
Please let this be a choice. One of the things I vastly prefer with opencode vs claude code is that I can type something to initiate the switch.. For example, sometimes the plan will just need a tiny tweak, and I can just write "begin implementation, but ensure that XYZ does {something slightly different to the plan}"
Vastly more token efficient than claude code and for me, a much more expressive and powerful workflow.
In my experience I feel the degrade (mostly minor) for this very reason because it causes an unwanted stir even with simple prompts to the build agent. It is only a "nice to have" for me, so not something breaking. If more people prefer what you said, the norm now should stick then.
IMO this should be a configurable permission on all agents, whether to ask, allow, or deny switching to other agents. This would greatly improve the ability to create custom agent workflows.
Activity
rekram1-node commentedon Nov 20, 2025
Yeah we are gonna fix this
TreeOfLearning commentedon Nov 21, 2025
Please let this be a choice. One of the things I vastly prefer with opencode vs claude code is that I can type something to initiate the switch.. For example, sometimes the plan will just need a tiny tweak, and I can just write "begin implementation, but ensure that XYZ does {something slightly different to the plan}"
Vastly more token efficient than claude code and for me, a much more expressive and powerful workflow.
aymuos15 commentedon Nov 21, 2025
In my experience I feel the degrade (mostly minor) for this very reason because it causes an unwanted stir even with simple prompts to the build agent. It is only a "nice to have" for me, so not something breaking. If more people prefer what you said, the norm now should stick then.
kylerm42 commentedon Nov 21, 2025
IMO this should be a configurable permission on all agents, whether to ask, allow, or deny switching to other agents. This would greatly improve the ability to create custom agent workflows.