Last Updated on August 24, 2026 by Editorial Team Author(s): Chew Loong Nian – AI ENGINEER Originally published on Towards AI. MCP Just Dropped 13 of Its 31 Methods. Why Did the Schema Still Grow 48%? Every summary of the 2026-07-28 Model Context Protocol release says the same thing in the same shape. The protocol went stateless. The initialize handshake is gone. Sessions are gone. Servers can run behind a plain round-robin load balancer now, or on a serverless function that forgets everything between invocations. The author digs into the claim that the protocol shrank by running an audit across three schema revisions, finding that while 13 of 31 methods truly disappeared, the schema actually grew by about 48% because the documentation—and required “envelope” metadata that replaces connection/session state—ballooned. They explain how the handshake functions were relocated into per-message metadata (including protocol version and client capabilities), why some features become opt-in or deprecated, how Tasks moved out of the core into an extension (with modern servers rejecting old task methods), and how subscriptions were rebuilt around client-driven listen streams instead of unsolicited server pushes. The piece further covers the new caching-oriented discovery call, new error types that reflect what the old handshake used to guarantee, and the practical cost trade-offs per request (larger payloads due to the envelope) versus operational benefits like stateless deployment and easier load balancing. It ends by recommending when to stay on legacy, serve both eras, go modern-only, or migrate urgently (especially if you relied on Tasks), with a practical “audit script” approach to measure the real bandwidth impact for your own workload. Read the full blog for free on Medium. Join thousands of data leaders on the AI newsletter. Join over 80,000 subscribers and keep up to date with the latest developments in AI. From research to projects and ideas. If you are building an AI startup, an AI-related product, or a service, we invite you to consider becoming a sponsor. Published via Towards AI
Watermarking Text Generation Efficiently