VS Code extension: add a verbose toggle equivalent to CLI Ctrl+O

Summary

In the CLI, Ctrl+O toggles between compact and verbose output, exposing thinking tokens and full tool inputs/outputs. In the VS Code extension, there is no equivalent: thinking is only shown as a transient Thinking… / Churning… status indicator that disappears as soon as the response begins, and tool inputs/outputs are not persistently expandable.

This makes it hard to follow what the agent is actually doing — especially during longer turns with multiple tool calls.

Current behaviour

• CLI: persistent verbose mode, toggled with Ctrl+O.
• VS Code extension: only an ephemeral status indicator. Thinking blocks are not user-expandable; tool I/O occasionally flashes by but cannot be reopened after the fact.

Workarounds users have today

• Run claude inside the VS Code integrated terminal to fall back to the CLI experience (loses the extension UI).
• Read the JSONL transcript files under ~/.claude/projects/<project-slug>/ after the session.

Both work, but neither is a substitute for an in-UI toggle.

Requested behaviour

• A toggle (e.g. a button in the panel header and/or a configurable keybinding) that switches the extension between compact and verbose views.
• In verbose view: thinking blocks rendered as expandable, persistent sections; full tool inputs and outputs available inline (collapsible).
• Ideally also exposed as a VS Code command (e.g. claudeCode.toggleVerbose) so users can bind it to their own keybinding (Ctrl+O is already taken by Open File in VS Code).

Why this matters

• Transparency into the agent's reasoning is one of the strongest reasons people enable extended thinking in the first place — the extension currently hides it.
• Debugging unexpected tool calls or wrong answers is much harder without seeing the chain of thought.
• It brings parity between CLI and extension users.

Environment

• Claude Code VS Code extension (latest at time of report)
• VS Code on Linux