Model Context Protocol updates target enterprise scale with stateless design

The Model Context Protocol (MCP), an open-source standard for how AI systems interact with external tools and data sources, has undergone its most significant update since its inception. The primary change is the transition of the protocol's core to a stateless design, meaning requests are no longer dependent on a session tied to a specific server instance. This aims to address previous barriers to scalability and improve reliability for MCP servers, a feature highly requested by developers.

This update, described as MCP's most important since remote MCP launched over a year ago, also introduces 'Multi Round-Trip Requests,' header-based routing, cacheable list results, enhanced authorization, a formal extensions framework, and updated Tier 1 SDKs. These enhancements are intended to make the protocol easier to deploy widely in enterprise environments. MCP, which began as a tool for connecting models to local applications, has undergone a significant architectural rethink.

Additionally, a new deprecation policy has been implemented, ensuring that at least 12 months will pass between the formal enactment of a feature's deprecation and its actual removal, with a narrow exception for critical security updates. This policy aligns with the overall goal of making MCP more robust for enterprise-scale use.

MCP is managed by the Agentic AI Foundation (AAIF), which operates under the Linux Foundation. Although originally introduced by Anthropic nearly two years ago and still significantly influenced by the company, MCP's development now includes contributions from OpenAI, Google, Microsoft, and Amazon, and it is supported by numerous developer tools and software services.