MCP stateful sessions force sticky load balancers and block horizontal scaling
The MCP spec requires a stateful initialization handshake. Putting an MCP server behind a standard load balancer breaks sessions because requests land on different instances. Operators must use sticky sessions (which fail on restart) or a shared Redis session store (adding a SPOF). This made horizontal scaling and Kubernetes-native deploys impossible without workarounds. The MCP spec team opened SEP-1442 in Sept 2025 to make the protocol stateless by default, which became the headline change in the 2026-07-28 RC.
A CLI tool that wraps stateful MCP servers and externalizes their session state so they run behind standard round-robin load balancers without sticky routing
173 โฒScore Breakdown
Social Proof 2 sources
Existing Solutions 2 competitors
MCP spec RC removes the stateful init handshake; servers can now sit behind round-robin load balancers
Shared Redis session store allows multiple instances to share MCP session state at cost of added complexity
Gap Assessment
MCP 2026-07-28 RC fixes this at protocol level but existing deployed servers must migrate; migration tooling is sparse