Worked on the pydantic-ai repository to extend the MCPToolset with background task support, enabling tools to declare and manage asynchronous tasks for improved workflow durability and cancellation. Focused on Python and asynchronous programming, the implementation introduced task-augmented execution, allowing for more resilient and flexible tool operations within MCP-based deployments. The approach emphasized robust API development and thorough testing to ensure reliability and alignment with the MCP roadmap. By co-authoring this feature, contributed to the SEP-1686 initiative, laying the groundwork for enhanced task management and more dependable automation in future releases without addressing bug fixes during this period.