The Model Context Protocol, or MCP, published a new version on July 28, 2026, that makes two practical changes: it stops servers from needing to remember details about each ongoing conversation, and it adds small labels to network traffic so that tools managing that traffic can identify it more easily. Both changes aim to make MCP servers easier to run when demand gets high.

To understand the first change, think about calling a company's customer service line. If the agent you reach has no memory of your previous calls, they can help you start fresh every time, and it doesn't matter which agent picks up. That is what "stateless" means. The old version of MCP worked more like a customer service line where you always had to reach the same agent who remembered your case. That made it hard to add more agents when call volume went up, because your call might land on someone who didn't have your context. The new version removes that requirement, so any server can handle any request. Arcade, a startup founded by Nate Barbettini, published a technical breakdown at arcade.dev/blog/mcp-going-stateless/ explaining the changes and what they mean for teams running MCP servers. (TechCrunch)

The second change requires MCP traffic sent over the web to include two small labels, called Mcp-Method and Mcp-Name, attached to each request. Think of these as envelopes that say what's inside before you open the letter. With these labels, network tools like load balancers and monitoring software can identify and route MCP traffic without having to open and read the actual contents of each request. The official MCP blog states these headers are intended to make the resulting traffic easier to operate. (MCP Blog)

TechCrunch, reporting on the update on July 20, 2026, describes MCP as "one of the basic building blocks of AI interoperability, giving AI models a secure way to access external data sources and services." The specification has been public since May, according to the same article, with the July 28 release candidate formalizing changes that had been under discussion in the interim. (TechCrunch)

Anthropic created MCP as an open standard for AI applications to connect to external tools and data. In December 2025, the company announced it was donating the protocol and establishing the Agentic AI Foundation as a neutral home for its governance. An official, community-driven MCP Registry was launched at the same time for discovering available MCP servers. (Anthropic)

The ecosystem around MCP has been expanding on several fronts. Reuters launched an MCP server on July 8, 2026, to deliver news content directly into AI workflows. On July 16, 2026, the Agentic AI Foundation introduced the MCPA, described as the first official certification for the Model Context Protocol. Arcade, which published the stateless transition walkthrough, raised $60 million in June 2026. (Reuters) (BusinessWire) (Agentic AI Foundation)

Anthropic's own engineering work has steadily pushed MCP into production use. A June 2025 engineering post described "Desktop Extensions," enabling one-click MCP server installation for Claude Desktop. A November 2025 post detailed code execution via MCP, explaining how the approach lets agents handle more tools while using fewer tokens and reducing context usage. Anthropic has also published support documentation for custom connectors using remote MCP and, in October 2025, added connectors for S&P Capital IQ and Daloopa as part of advancing Claude for Financial Services. In August 2025, Anthropic and Google Cloud co-presented a webinar on deploying multi-agent systems using MCP alongside the Agent-to-Agent (A2A) protocol with Claude on Vertex AI. (Anthropic Engineering) (Anthropic Engineering) (Anthropic Support) (Anthropic) (Anthropic Webinars)

For anyone who has dealt with servers that need to remember session details, the pain is familiar. If the server holding your session crashes, your session is gone. Adding more servers to handle more traffic becomes complicated because each user has to stay connected to their original server. MCP servers handling high volumes of AI requests face the same constraints. Removing server-side session memory means a request can land on any healthy server, which simplifies things when servers fail or when traffic spikes. The new traffic labels complement this by giving network tools enough information to route and monitor MCP traffic without inspecting its contents.

In my view, these two changes lower the real-world cost of running MCP infrastructure at a moment when adoption is accelerating. Reuters building an MCP server, the Agentic AI Foundation introducing certification, and Arcade securing funding to build tooling around the protocol all point to MCP moving from specification toward production standard. The protocol's donation to a foundation and the launch of a community registry, both in December 2025, established the governance scaffolding. This release candidate does the quieter, less glamorous work of making the plumbing reliable enough that infrastructure teams will adopt it without caveats.

For engineers currently building MCP servers that keep session state, the release candidate is worth evaluating against existing deployment patterns. The stateless model shifts responsibility for any required state to the client or to an external store, which is the standard tradeoff. The header changes are lower-risk but require updates to transport-layer code. Both are backward-incompatible and will require coordinated client and server updates.