Quick Facts

  • The 2026-07-28 MCP specification introduces a stateless protocol core, replacing the previous session-based handshake workflow that created scaling problems in cloud deployments.
  • Six Specification Enhancement Proposals harden the authorization layer against OAuth mix-up attacks and align the protocol more closely with OAuth 2.0 and OpenID Connect standards.
  • SDK downloads have reached roughly 250 million per week, doubling in the past six months, according to MCP co-creator David Soria Parra.

The Model Context Protocol has its largest update since Anthropic launched the open standard in late 2024. The 2026-07-28 specification revision delivers a stateless architecture, new security controls, and a formal extensions framework designed to move agentic AI into large-scale enterprise production environments.

MCP is an open standard that gives AI applications a single, uniform interface to connect with external tools and data. When Anthropic released the first version in 2024, developers used it primarily on personal computers. As those deployments moved into the cloud and scaled across multiple clients, the original session-based design created bottlenecks.

Stateless at the Core

The headline change in the new spec is the removal of the stateful handshake workflow MCP previously used to transmit request metadata. A remote MCP server that once required sticky sessions, a shared session store, and deep packet inspection at the gateway can now run behind a plain round-robin load balancer. Clients can cache tools and list responses for as long as the server’s ttlMs parameter permits, and traffic routes on a standard Mcp-Method header.

“A lot of things that made MCP are gone,” said David Soria Parra, MCP co-creator and lead maintainer at Anthropic. “This is a protocol growing up in real time around what production teams actually need.”

Authorization and Security Hardening

Six Specification Enhancement Proposals address the authorization layer. Clients must now validate the iss parameter on authorization responses per RFC 9207, a requirement that closes a class of OAuth mix-up attacks more common in MCP’s single-client, many-server deployment pattern.

Additional changes include application type declaration during Dynamic Client Registration, fixing failures where authorization servers defaulted desktop or CLI clients to a web application type and rejected localhost redirect URIs. Credential binding now ties registered credentials to the issuing authorization server, with re-registration required when a resource migrates between servers.

Dynamic Client Registration is formally deprecated in favor of Client ID Metadata Documents. It remains functional for backward compatibility but will be removed in a future version of the spec.

The Enterprise-Managed Authorization extension went stable on June 18, 2026. It replaces per-server OAuth consent prompts with a single enterprise login already supported by seven MCP servers.

A Formal Extensions Framework

The update ships a structured extensions framework that gives new capabilities a defined path from experimental to official status. Extensions use reverse-DNS identifiers, are negotiated through a capabilities map, and version independently of the core specification.

Two official extensions launch with the new spec. MCP Apps allows servers to deliver interactive HTML interfaces that MCP hosts render in a sandboxed iframe. The Tasks extension supports long-running agent workflows through a server-directed lifecycle and handle-based state management.

Adoption and Enterprise Readiness

Soria Parra described the update as the clearest signal yet that MCP is becoming production-grade infrastructure. He called the download figures, roughly 250 million SDK downloads per week, “just insane numbers.” The figure represents a doubling over six months.

For software leaders evaluating agentic AI infrastructure, the stateless core reduces operational overhead significantly. Scaling no longer requires specialized gateway configurations or distributed session stores, which lowers both the cost and complexity of deploying MCP-based systems at enterprise volume.

Read more: Biggest ever MCP update brings metadata, cybersecurity enhancements

ai agentsanthropicapicybersecurityenterprise

This article was written by an AI agent. Spotted an error? Send a correction and we will fix it.