MCP publishes its next roadmap

The Model Context Protocol’s Core Maintainers published an updated roadmap on August 22, covering the next specification release and beyond, developed with the community and the protocol’s Working Groups. The post frames it as a checkpoint after the 2026-07-28 release, which rewrote MCP’s core to be stateless, added Multi Round-Trip Requests, header-based routing, and cacheable list results. Tier 1 SDKs are now seeing close to half a billion downloads a month, with the TypeScript and Python SDKs each past a billion total downloads.

Five priorities are named for what ships next: server-initiated events (webhooks and channels) and a maturing Tasks extension for agentic messaging; extending Streamable HTTP over stdio so local and remote servers develop the same way; agent identity built on Demonstrating Proof of Possession and Workload Identity Federation, aimed at letting agents delegate to each other without sharing API keys; standardized tool-result handling and progressive tool discovery to cut the context overhead of large tool lists; and broader SDK ergonomics and conformance testing. Governance also formalized a Contributor Ladder and gave Working Groups authority to triage proposals in their own areas.

What it means for you

If you’re running MCP servers in production, the near-term payoff is in the identity and tool-discovery work: DPoP-based agent identity removes a real operational headache (shared long-lived API keys between agents), and progressive tool discovery directly addresses the context bloat that shows up once a server exposes more than a handful of tools, a problem the July stateless rewrite already forced most server maintainers to confront. None of this is shipped yet, but it tells you where to expect the next breaking-change wave.