Version 2

The 2026-07-28 MCP specification release candidate redesigns the protocol's transport layer to be stateless, removing per-session, per-server-instance dependencies that had made standard cloud load-balancing incompatible with MCP [1][2]. Microsoft's Azure App Service team has since published scaling guidance based on the new spec [9], adding a major cloud provider voice to what had been primarily developer and analyst coverage. Commercial advisory content on enterprise MCP adoption practices is accumulating from multiple outlets [11][12], while technical breakdowns of the spec's migration implications [5] and developer community discussion of remaining barriers [19] continue in parallel.

MCP's stateful session model was a concrete obstacle to enterprise cloud deployment. The stateless redesign removes that constraint, and Microsoft publishing Azure App Service guidance suggests that cloud infrastructure teams are treating the spec as production-relevant — not just a developer experiment. Whether remaining friction points (setup complexity, cross-platform divergence, and challenges the spec doesn't address) slow actual adoption is the live question.

The 2026-07-28 MCP specification release candidate introduces a stateless protocol core, the most significant technical change to the Model Context Protocol since Anthropic introduced it [1]. The spec was authored by Anthropic's David Soria Parra and Den Delimarsky, the protocol's lead maintainers. Under the prior design, MCP sessions were bound to individual server instances, which made standard cloud load-balancing — routing requests across a pool of servers — incompatible with the protocol. The new spec removes that constraint: requests carry no session affinity, so any server instance can handle any request [1][2][3]. Ars Technica and VentureBeat both frame this as the largest update the protocol has received [1][4]; InfoWorld focuses on the scaling implications [3]; Stacktree has published a technical breakdown of what specifically changed and what existing implementations may break [5].

The stateless redesign fits within a broader 2026 roadmap for MCP emphasizing enterprise readiness [6][7]. TheNewStack identifies the prior stateful model as the main barrier to enterprise adoption and frames the new spec as its resolution [8]. Microsoft's Azure App Service team has published guidance on deploying MCP under the new stateless model [9], adding a major cloud provider voice to what had been primarily developer and analyst coverage. A commercial advisory layer is also forming: Manufact (YC S25) launched an MCP Cloud product [10], CData and Nordic APIs have published enterprise adoption guides [11][12], and Gainsight has documented an MCP server integration for its customer success platform [13].

Multi-platform support has arrived in parallel. Both Claude and ChatGPT now accept custom MCP server connections [14][15][16]. Simon Willison documented the process as feasible but involving 'quite a few steps' [14], and Docker has published guidance on adding MCP servers to ChatGPT [16]. Community developers are actively experimenting with cross-platform MCP setups connecting Claude, ChatGPT, Gemini, and Perplexity [17]. Hacker News, Reddit, and the Ars Technica forum show active developer engagement with the spec [2][18][19][20].

Not all observers read the stateless spec as a complete solution. A Medium analysis identifies six critical challenges still facing MCP in 2026, arguing the spec is a meaningful step without being a full answer to enterprise adoption barriers [21]. Developer community discussion on Reddit surfaces practitioner-level concerns about enterprise adoption that go beyond the transport layer [19]. Developer friction in setup and the divergence between Claude and ChatGPT implementations remain open practical questions. The specification is currently a release candidate, so final ratification and client-side implementation timelines are not yet confirmed.