01Multi-level reasoning (basic, normal, high) with configurable thought counts and token budgets
02Optional task execution with progress notifications for multi-step requests
03Dual execution modes: 'step' for single thoughts and 'run_to_completion' for batched thoughts
04Resource endpoints for viewing session lists, details, markdown traces, and individual thoughts
05Prompt and resource argument completion for seamless interaction
060 GitHub stars