A remote MCP server that lets an AI assistant session and a separate agent exchange queued messages via tools like send_to_chief and wait_for_reply, waking the agent with webhooks and long-polling for its replies.