The MCP protocol is undergoing its biggest update since its launch 16 months ago, involving a major rewrite aimed at simplifying its core. Originally designed as a stateful protocol requiring handshakes and session IDs, MCP faced significant production challenges, especially in large-scale deployments. These challenges included the need for sticky sessions, complex infrastructure like Redis clusters, and incompatibility with serverless architectures. To address these issues, the new version removes these stateful elements, making the protocol stateless so that any server instance can handle any request, enabling easier load balancing, auto-scaling, and serverless deployment.
Despite its rapid adoption—boasting 110 million SDK downloads per month—MCP has faced criticism and skepticism from the developer community. Some have abandoned the protocol in favor of simpler CLI tools and skills, citing overhead and complexity. Moreover, readiness for the new update is low, with only one out of over 4,000 servers ready to support the new spec two weeks before release. This highlights the challenge of rolling out such a fundamental change across a widely deployed standard.
The rewrite also shifts state management responsibility from the protocol to the application layer. Instead of the server maintaining session state, the client now carries a sealed state blob between requests, allowing any server instance to resume work. While clever, this approach places the burden of state tracking and security on developers, raising concerns about potential authorization vulnerabilities, especially since MCP’s authorization mechanisms remain fragmented and inconsistent across clients. The update includes some hardening of authorization but does not unify or fully resolve these issues.
Despite these challenges, MCP’s widespread adoption in enterprise environments like Salesforce, Jira, and Snowflake underlines its importance and staying power. The protocol’s rapid growth outpaces many other standards, and its integration behind corporate firewalls ensures it remains a critical infrastructure component. Some complaints, such as context bloat, have been addressed through client-side optimizations rather than protocol changes, showing that MCP’s roadmap continues to evolve pragmatically.
In conclusion, MCP’s rewrite is a survival strategy acknowledging past design mistakes and simplifying the protocol to ensure long-term viability. While the update may cause short-term disruption and requires server maintainers to adjust their implementations, the protocol remains the fastest-adopted AI-era standard with no clear rivals. The success of this transition will depend on whether the ecosystem can unify around improved authorization and whether more servers adopt the new spec in the coming months. MCP deliberately made itself “dumber” to become more robust and adaptable in the complex, large-scale environments it now serves.
