Problem
Plan mode works well for complex tasks: the agent researches the codebase, asks questions, and produces an editable plan before any implementation.
Today, clicking Build is the intended approval step, but it also switches the session out of Plan mode into Agent-style execution. The mode indicator changes, and the workflow feels like “plan in one mode, code in another” rather than one continuous plan-driven session.
I would like Build to mean “approve and execute the plan” while the chat remains in Plan mode — same mode picker state, same mental model: I am still working from the plan, not in a separate Agent session.
Expected behavior
User creates and edits a plan in Plan mode.
User clicks Build to approve execution.
The agent implements the plan (files, terminal, TODO progress) in the same chat.
The mode picker stays on Plan for the entire session (planning + execution).
Plan TODOs and Built status continue to track progress as today.
Why this matters
Clear separation: Plan = plan-driven work end-to-end; Agent = ad-hoc implementation without a formal plan step.
Reduces confusion when Build feels like “switching products” mid-task.
Easier to iterate: refine plan → Build → partial result → adjust plan → Build again, without mode churn.
Aligns with users who want explicit approval (Build) but not a mode change.
Current workaround (unsatisfactory)
Keep a Plan chat without Build and a separate Agent chat with the plan pasted in — loses session linkage and Built tracking.
Use Build and accept Agent mode — works functionally, but not the UX I want.
Suggested implementation (optional)
Setting under Settings → Agents, e.g. “Execute plan in Plan mode (do not switch to Agent after Build)”.
Or: Build always runs plan execution in Plan context; Agent mode remains for tasks started without a plan.
Hey, thanks for the detailed feature request. It’s well written, and I get what you’re aiming for.
Your request for Build to execute the plan without switching into Agent mode, as one continuous plan-driven session, is noted. It overlaps with a couple existing threads about mode churn that you linked, so it’s not a one-off, which helps with prioritization.
I can’t share an ETA or make any promises on implementation. If there are updates, we’ll post them here. I’ll keep the thread in Feature Requests so others can upvote.
It would also be great to have the Build in New Chat button to start executing a plan in a fresh context, rather than continuing building in the same half-clogged chat that was used for planning
Hey @Oleg_Mastriukov, good news, this already works in the IDE as Build in New Agent. It’s a bit hidden: open the plan in the dynamic .md view, not raw markdown, select at least one todo, and you’ll see the Build in New Agent button. It runs in a fresh agent context, so it doesn’t carry over the messy planning chat.
This button isn’t available in the Agents Window yet. But in the normal editor layout like in your IDE, it should be there.
Let me know if the button doesn’t show up, then we can check your version and see what’s blocking it.
