Server-Controlled Extended Thinking (Experimental)

What: Server-side feature flag system that enables automatic extended thinking token allocation for specific users without requiring manual triggers.

Details:

• New API endpoint /api/oauth/claude_cli/client_data fetches user-specific configuration
• When alwaysThinking is enabled server-side, maximum thinking tokens are automatically allocated
• Thinking blocks are intelligently hidden when auto-enabled unless user explicitly requested thinking
• Does not affect manual thinking triggers or MAX_THINKING_TOKENS environment variable
• Evidence: Client data API in TWA() at line 361888, token allocation in pT() at line 362093, hiding logic at line 415872