Learn about the architectural decisions and implementation challenges behind Anthropic's remote Model Context Protocol (MCP) support in this 15-minute conference talk by software engineer John Welsh. Discover how Anthropic integrated remote MCP functionality into both claude.ai and the Anthropic API, exploring key technical aspects including authentication mechanisms, support strategies for engineers developing agentic AI tools, and custom internal transport implementations. Gain insights from Welsh's 20 years of experience building large-scale systems as he shares practical lessons learned from shipping this significant feature, covering the technical decisions that enable seamless communication between internal systems and MCP servers along with other critical integrations.