This week, the Model Context Protocol (MCP) - an open source standard for how AI systems talk to external tools and data sources - got its biggest update since it first showed up. The headline change: MCP’s protocol core is now stateless, meaning requests no longer need to cling to a specific server session. Translation: it might finally stop acting like a clingy ex when you try to scale it up.

The blog post from lead maintainers David Soria Parra and Den Delimarsky (both of whom happen to work at Anthropic) says: “The highlight of this release is a stateless protocol core - MCP is transforming from a bidirectional stateful protocol into a request/response stateless protocol. It was one of the most highly-requested features from developers who were eager to get better reliability and scalability for their MCP servers.”

Soria Parra also called this “MCP’s most important since remote MCP first launched over a year ago.” The update also adds multi round-trip requests, header-based routing, cacheable list results, authorization hardening, a formal extensions framework, and updated Tier 1 SDKs. Basically, a whole lot of jargon that hopefully adds up to “works better for big companies.”

MCP started life as a local-machine-only affair, connecting models to local apps, so this spec represents a major rethink. There’s also a new deprecation policy giving at least 12 months’ notice before features get the axe (unless there’s a critical security bug, in which case all bets are off). All very enterprise-friendly.

MCP is now managed by the Agentic AI Foundation (AAIF), under the Linux Foundation. Anthropic originally introduced it almost two years ago and still has significant influence, but OpenAI, Google, Microsoft, and Amazon also chip in. It’s supported by loads of developer tools and increasingly by software for knowledge and creative work. The buck stops with individual maintainers - some of whom, again, are from Anthropic - but hey, no conflicts of interest here, right?