Glonce

MCP goes stateless in its largest update since launch

infrastructureThe Glonce editorsArs Technica🔥77Glonce rating77Impact70Novelty75Relevance100Surprise55Players65Credibility85

The Model Context Protocol (MCP), the open source standard that lets AI systems talk to external tools and data sources, has received its largest update since it launched. The change at the center of the release: MCP's protocol core is now stateless. Previously the protocol was bidirectional and stateful, meaning a request depended on a session tied to one specific server instance. That dependency is gone; MCP now works as a request/response protocol, so requests no longer need to stick to a particular server. The announcement, posted on the official MCP blog by lead maintainers David Soria Parra and Den Delimarsky, both of whom work at Anthropic, frames the statelessness as the release's headline feature and says it was one of the most requested changes from developers, who wanted better reliability and scalability for their MCP servers. Soria Parra separately called it MCP's most important update since remote MCP first launched, over a year earlier. Beyond the stateless core, the update bundles several other changes: Multi Round-Trip Requests, header-based routing, cacheable list results, hardened authorization, a formal extensions framework, and updated Tier 1 SDKs. Ars Technica frames the whole package as targeting the main obstacle to running MCP at enterprise scale, since session-bound requests made it hard to scale MCP servers horizontally or run them reliably at scale.

Key facts

  • MCP's protocol core switched from a bidirectional, stateful design to a stateless request/response model, the largest change to the specification since MCP's introduction.
  • The old design tied every request to a session on one specific server instance; that dependency is what the stateless core removes, targeting the main barrier to running MCP at enterprise scale.
  • The update was announced by lead maintainers David Soria Parra and Den Delimarsky, both at Anthropic, who call the stateless core one of developers' most requested features for better reliability and scalability.
  • Soria Parra separately described the release as MCP's most important update since remote MCP launched, over a year earlier.
  • The same release also adds Multi Round-Trip Requests, header-based routing, cacheable list results, authorization hardening, a formal extensions framework, and updated Tier 1 SDKs.

Why it matters

MCP has become the default way AI systems plug into external tools and data, and its stateful design was a known ceiling: every request had to stay pinned to the server instance that started the session. That ruled out standard ways of scaling MCP servers reliably. Moving the protocol core to stateless request/response removes that constraint at the specification level, which is why the maintainers call it the release developers asked for most.

Who it affects

Anyone running or building on MCP servers: enterprises deploying MCP at scale, who hit the session-affinity ceiling directly; developers maintaining MCP server implementations, who now need to track the new request/response model; and SDK users, since Tier 1 SDKs were updated alongside the spec. Anthropic, which employs both lead maintainers, remains the protocol's primary steward.

How to use it

Teams operating MCP servers should move to the updated Tier 1 SDKs and take advantage of the new capabilities the same release bundles: Multi Round-Trip Requests, header-based routing, cacheable list results, a hardened authorization model, and a formal extensions framework for building on top of the spec.

How solid is it

The information comes straight from the MCP maintainers' own announcement, written by Soria Parra and Delimarsky, who lead the specification and work at Anthropic, the protocol's main backer, giving it primary-source weight on what changed and why. It is a first-party framing of the release's intent, though; independent confirmation of the reliability and scalability gains in production would come from how MCP server operators fare after adopting the update.

Risks and caveats

The available material does not give a version number, a rollout timeline, or a migration path for MCP servers already running on the old stateful model, so backward compatibility and the practical cost of upgrading existing deployments are open questions. The claim that this fixes enterprise scaling is the maintainers' own characterization; whether it holds up depends on adoption still to come.

“It was one of the most highly-requested features from developers who were eager to get better reliability and scalability for their MCP servers.”

— David Soria Parra and Den Delimarsky, MCP lead maintainers