Model Context Protocol prepares to break with its stateful past
The Model Context Protocol (MCP), an open standard for linking AI applications to external systems, is due for its largest overhaul since launch. Its maintainers plan to finalise a new revision on 28 July that removes protocol-level sessions and other features, aiming to make cloud deployment and scaling simpler but creating compatibility work for organisations with custom implementations.
The change reflects problems encountered as MCP moved from laptop-based demonstrations to multi-client enterprise services in the cloud. Under the new stateless design, each request carries version, client identity and capability information in its metadata, avoiding session tracking; MCP’s SDK had exceeded 97 million monthly downloads and at least 10,000 servers existed when Anthropic transferred the protocol to the Linux Foundation-backed AAIF in 2025.
In short
- MCP will become stateless to simplify enterprise scaling.
- Custom MCP implementations may require substantial reworking.
- The revision is scheduled for finalisation on 28 July.
Steelman: both sides, in good faith
The strongest fair case each way — we don't pick a winner.
The case for
The move to a stateless architecture addresses real scalability challenges that emerged only after MCP moved from development demonstrations to large-scale enterprise deployment. Stateless designs are a proven pattern for cloud systems, simplifying load balancing, horizontal scaling and operational management. With 97 million monthly downloads and thousands of operational servers, the protocol's maintainers are justified in prioritising these architectural improvements, even if implementations must update to accommodate the change.
The case against
Protocol breaking changes impose substantial costs on organisations with custom implementations, forcing them to undertake significant compatibility work that disrupts their operations and diverts resources. Rather than shifting this burden to users, MCP's maintainers should have maintained backward compatibility or provided smoother migration paths, allowing organisations to upgrade at their own pace rather than facing a forced architectural overhaul.
Explained simply
The Model Context Protocol is an open standard that allows artificial intelligence applications to connect with external systems and data sources. Originally developed by Anthropic, the protocol moved to a Linux Foundation-affiliated organisation in 2025 as adoption grew significantly. It is now widely used, with its software development kit downloaded over 97 million times and thousands of organisations running their own servers.
The protocol is undergoing its most significant revision since creation. The update will redesign how it tracks information, moving to a system where each request contains all necessary details rather than relying on ongoing sessions. This change aims to simplify how organisations deploy and scale MCP services in cloud environments.
The overhaul reflects challenges encountered as MCP evolved from simple demonstrations on individual computers to complex enterprise systems serving multiple clients simultaneously. Whilst the new design will ease scaling for most users, organisations that have built custom implementations of the protocol will need significant compatibility work to align with these changes.