AI’s most important protocol is getting a little bit easier to use

The Model Context Protocol (MCP) is a fundamental building block for AI interoperability, enabling AI models to securely access external data sources and services. It acts as plumbing that allows a chatbot to reach into a calendar, database, or internal tools without engineers having to build custom connections for each one. Next week, the protocol is getting a significant update that, while invisible to end users, could have a big impact on the ecosystem’s development. The official spec for the new version has been public since May, but on Monday, Arcade—a two-year-old startup built around making AI agents function inside real companies—provided an unusually clear explanation of the changes. Arcade raised $60 million in June on the premise that most AI agents fail not because of weak models but because of inadequate infrastructure, and this update aims to address that. Specifically, MCP is changing how it handles session IDs—tokens that servers use to recognize the same conversation. Under the current system, when a client like Claude first connects to a server, it sends a greeting with its version and capabilities. The server responds with its own capabilities and a session ID. From then on, the client sends that ID with every request. Sometimes the ID expires, and the client must request a new one. In real deployments with millions of users and load balancers distributing requests across many servers, this creates headaches because every server must know about session IDs issued by others. The new approach makes session IDs stateless on the server side, similar to how most ordinary websites work. This should make the system easier to maintain and potentially cheaper to run at scale. While some parts of AI development move rapidly, this update is a reminder that standards and infrastructure progress more slowly. It is happening, but at a slower pace.