The `BaseSession` class in the `core/mcp/session` package uses `ThreadPoolExecutor` to run the receive loop but fails to properly clean up the executor and receiver future, leading to potential thread leaks. This PR addresses this issue by: - Initializing `_executor` and `_receiver_future` attributes to `None` for proper cleanup checks - Adding graceful shutdown with a 5-second timeout in the `__exit__` method - Ensuring the ThreadPoolExecutor is properly shut down to prevent resource leaks This fix prevents memory leaks and hanging threads in long-running scenarios where multiple MCP sessions are created and destroyed. Signed-off-by: neatguycoding <[email protected]> Co-authored-by: QuantumGhost <[email protected]> Co-authored-by: Copilot <[email protected]>