MCP just got its biggest update ever — here’s what changes for AI agents
5:00 am, PT, July 28, 2026
The Model Context Protocol, the open standard that has quietly become the connective tissue between AI agents and the world’s software, is getting its largest update since Anthropic released it twenty months ago — a sweeping architectural revision that its maintainers and backers say finally makes agentic AI ready for massive enterprise production deployments.
The update, released today under the stewardship of the Agentic AI Foundation (AAIF), a directed fund under the Linux Foundation, finalizes MCP’s transition to a fully stateless architecture, hardens its authentication model against a known class of attacks, establishes a formal 12-month deprecation policy, and graduates two headline capabilities — interactive server-rendered interfaces and long-running asynchronous tasks — into official protocol extensions.
According to the announcement, running MCP at scale has historically required “sticky routing” or shared state to maintain continuity across sessions. The new release removes that bottleneck, letting organizations run MCP servers behind standard load balancers using Kubernetes and cloud-native DevOps tooling.
“Some people jokingly call it a v2, and I think in spirit that’s accurate,” David Soria Parra, MCP’s co-creator and a lead maintainer at Anthropic, told VentureBeat. “It’s probably the biggest change we’ve ever made to the protocol, and with that, it’s a big step up in maturing it for use by really big players.”
Why stateless architecture matters
Under the old design, an MCP client had to maintain a persistent session with a specific server instance. In modern cloud environments that was a major operational burden.
“Before, you needed to have a session store and manage session IDs — and if one of your compute pods went down, all of a sudden the requests would start failing,” said Den Delimarsky, a lead maintainer of the protocol. “That’s not going to be a problem with the new version of the protocol.”
Mazin Gilbert, executive director of the AAIF, compared the change to the architectural decision that made the web itself possible: clients can speak to any server behind a load balancer without sticky sessions.
Trade-offs include larger payloads (state moved onto the wire) and deprecation or narrowing of rarely used capabilities such as out-of-band server logging. Application state is now managed via explicit handles passed as tool arguments rather than protocol-level sessions.
Deprecation policy, auth hardening, Apps and Tasks
The release adds a formal deprecation framework guaranteeing at least twelve months between deprecation and earliest removal — consulted with Google, Microsoft, and Amazon.
Authorization hardening aligns MCP with OAuth 2.0 / OpenID Connect practice, including mandatory validation of the issuer (iss) parameter to close mix-up attacks. An Enterprise Managed Authorization extension was developed with Okta.
Two capabilities graduate to official extensions: MCP Apps (server-rendered interactive UIs in sandboxed iframes) and MCP Tasks (durable task handles for long-running work so clients can disconnect and resume polling). Multi-round-trip requests allow mid-call elicitation without holding SSE streams open.
Scale and governance
Anthropic created MCP in November 2024 and donated it to AAIF under the Linux Foundation in December 2025. AAIF membership grew from ~40 at inauguration to ~240. SDK downloads are reported at roughly 250 million per week.
Official specification reference: https://blog.modelcontextprotocol.io/posts/2026-07-28-release-candidate/ — final MCP 2026-07-28 ships July 28, 2026, replacing 2025-11-25. Breaking changes include removal of initialize/initialized handshake and Mcp-Session-Id (SEP-2575, SEP-2567). Roots, Sampling, and Logging are deprecated under the new lifecycle policy.