MCP’s Stateless Session Update Will Make AI Agents Easier to Scale

The Model Context Protocol (MCP), a key building block for AI interoperability, is getting an update next week that aims to simplify the way servers handle session IDs. While invisible to end users, the change could have significant implications for developers running MCP servers at scale.

Currently, MCP uses a stateful session model: when a client (like Claude) first connects to a server, it performs a handshake and receives a session ID. Every subsequent request includes that ID so the server can maintain context. But this design becomes problematic behind a load balancer, where requests may be routed to different server instances. Each machine must know about session IDs issued by others, adding complexity and overhead.

Arcade, a startup that raised $60 million in June to build infrastructure for AI agents, explained the update. Founder Nate Barbettini noted that the new version takes a “stateless” approach to session IDs on the server side, similar to how most ordinary websites work. This makes it much easier to run MCP servers behind load balancers and scale horizontally.

The official spec for the new version has been public since May, but the upcoming release marks a practical step forward. For developers, this means less custom work to maintain session state across distributed systems, and potentially lower costs for running MCP servers. It’s a reminder that while model capabilities advance rapidly, the infrastructure layer often evolves through slower standards processes—but it is moving.