The Model Context Protocol (MCP) — the open standard for connecting AI models to tools and data — is heading toward its biggest revision since launch.
- A release candidate for the 2026-07-28 MCP specification is out now, with the final spec expected to ship July 28, 2026 (note: that’s later than today, so treat this as an upcoming release still in review, not something already final).
- The headline change is a stateless core that scales on ordinary HTTP infrastructure. In practice, a remote MCP server that previously needed sticky sessions, a shared session store, and deep packet inspection at the gateway will be able to run behind a plain round-robin load balancer, route traffic using an
Mcp-Methodheader, and let clients cachetools/listresponses for as long as the server’sttlMsallows. - New extensions are part of the package: MCP Apps for server-rendered UIs, and a Tasks extension for long-running work that doesn’t block on a single request/response cycle.
- Authorization is being aligned more closely with standard OAuth and OpenID Connect deployments, which should simplify enterprise integration considerably.
- Looking further out, the roadmap includes event-driven triggers (webhooks), streamed and reference-based results, and a Skills primitive for composed capabilities — allowing servers to expose reusable, higher-level “skills” rather than just flat tool lists. This is still a roadmap item, not a shipped feature.
Why it matters: MCP has been widely adopted since its introduction, but production deployments have run into real friction around session management, streaming, and auth. This overhaul directly targets those pain points, and the upcoming Skills primitive could reshape how tool servers are structured going forward.