[FEATURE] Agentic mode toggle (like plan mode) to control subagent spawning

Preflight Checklist

• [x] I have searched existing requests and this feature hasn't been requested yet
• [x] This is a single feature request (not multiple features)

Problem Statement

Claude Code can spawn subagents automatically to parallelize independent tasks — but this happens transparently without user awareness or consent. This has two consequences:

1. Token usage: Each subagent has its own context window, so parallel execution multiplies token consumption without the user realizing it.
2. Loss of control: Users who want to control execution behavior have no way to opt out of subagent spawning short of hoping it doesn't happen.

Proposed Solution

Add an agentic mode toggle, analogous to how plan mode works, that controls whether subagents can be spawned automatically:

• Off (default): No subagents are spawned. All work happens inline in the main context.
• On: Current behavior — subagents can be spawned to parallelize independent tasks.

Users can explicitly enable it when they want faster parallel execution, and stay in control otherwise.

Alternative Solutions

• Always notify the user before spawning a subagent and require confirmation (more granular but more disruptive)

Priority

Medium

Feature Category

Core CLI behavior / execution control

Use Case Example

A user working on a large codebase asks Claude Code to make changes across multiple files. Without agentic mode awareness, subagents spin up silently, consuming significantly more tokens than expected. With an agentic mode toggle, the user can decide upfront whether they want parallel execution or prefer sequential, predictable, token-efficient behavior.