Uses a small LLM (Qwen) as a context bridge to enhance prompt with context for models that don't support MCP natively