Is your feature request related to a problem? Please describe.
I would like to be able to horizontally scale my MCP server while using streamable HTTP. Right now, it seems like there is a "pick-your poison" scenario: As a MCP server author I can either

• Use stateful mode, but give up on horizontal scalability,
• or use stateless mode and give up on sampling, elicitation, and progress reporting.

Describe the solution you'd like
I would like to be able to horizontally scale my MCP server while using stateful streamable HTTP. For that, I would envision something like socket.io's Redis backend , where all state is stored, such that everything "just works".

Describe alternatives you've considered
Stateless streamable HTTP doesn't work for me, because I would need to give up on features such as solicitation, sampling and progress reporting.

Additional context
Horizontal Scalability is important because it increases the resiliency and availability of the system - as demand grows, so does the number of replicas of the service. Unhealthy replicas are killed and replaced by healthy ones. In front of the service sits a load balancer that is aware of the replicas and routes requests to them.

Some applications have internal state that needs to be persisted across requests (e.g. PHP's $_SESSION state). For that, most people use a solution like Redis or ValKey, such that state is efficiently stored in a cache, as opposed to in-memory for each replica. For examplesocket.io, a WebSocket abstraction library, that manages potentially long-living connections, also supports using redis as a session state backend.