344
Productivity & Workflow355
Automation & Workflow224
Software Development251
Marketing & Growth192
AI Infrastructure & MLOps174
Writing & Content Creation203
Data & Analytics141
Photography & Imaging156
Design & Creative170
Customer Support131
Sales & Outreach125
Voice & Speech135
Education & Learning131
Operations & Admin87
The Model Context Protocol got a major update that removes server sessions and adds a clearer policy for removing old features.
In short: The Model Context Protocol (MCP), a shared rulebook for how AI connects to tools and data, has released its biggest update yet, and its core now works without server sessions.
MCP is an open source standard, meaning it is a public set of instructions anyone can use and improve. It helps AI systems connect to outside tools and data sources, like a plug that lets an app talk to other services.
This week, MCP’s maintainers announced a major change. The protocol core is now “stateless.” In simple terms, each request can be handled on its own without relying on a special ongoing session tied to one specific server (like being able to talk to any customer service agent, not the exact one you spoke to last time).
The update also adds features aimed at larger, more complex setups. These include multi round-trip requests (a way to handle back and forth steps), header-based routing (labels that help direct requests to the right place), cacheable list results (saving some responses so they can be reused), stronger authorization (tighter rules around who is allowed to do what), a formal extensions framework (a defined way to add optional extras), and updated software toolkits for developers.
MCP also introduced a new deprecation policy. It says there should be at least 12 months between officially warning that a feature will go away and actually removing it, except for urgent security fixes.
MCP is managed by the Agentic AI Foundation under the Linux Foundation. It began at Anthropic, and companies including OpenAI, Google, Microsoft, and Amazon also contribute.
If more companies want AI systems that can safely and reliably connect to business tools, they need standards that scale without breaking. A stateless design can make it easier to run many copies of a service at once and keep it dependable during busy periods.
Source: Arstechnica