Autonomous code-debugging agent that doesn't just suggest fixes, it verifies them. It uses a RAG-powered context engine to understand your codebase, an orchestrator loop to plan the fix, and an execution engine that actually runs the fix in a sandboxed environment. Then a critic agent challenges the output before anything gets committed. It's the full predict → decide → act loop, with an observability dashboard so you can see why it made every decision.