Tue, 3 Nov 2026
Programming

MCP 2026-07-28 introduces a stateless core and formal extensions framework

The new specification changes how AI applications communicate with tool servers. Each request carries its protocol context, making client–server compatibility and complete user journeys central to implementation testing.

Osamah Ismail · Founder & Editor-in-Chief
· 2 min read · Updated:
MCP 2026-07-28 introduces a stateless core and formal extensions framework
Official identifying artwork from the Model Context Protocol blog; not a test result for a particular system.

The maintainers of Model Context Protocol released MCP 2026-07-28 on July 28, 2026. Its central change is a stateless core: each request carries the information needed for the server to process it independently.

The release matters to teams connecting AI applications with tools and data. It changes the communication model, making the protocol versions supported by both client and server important implementation details.

What an independent request means

The versioned specification says a server must not infer a new request’s context from earlier requests. Information such as protocol version and client capabilities accompanies each request.

An application can still manage work across multiple steps through an explicit identifier. In an illustrative workflow, a tool creates a draft and returns its identifier. The next request includes that identifier to continue working on the same draft.

This makes the intended object visible at each step. Developers can examine its identifier together with the operation’s result when reviewing the task’s execution.

A useful exercise could include several independent drafts or objects. The team can then check that each request reaches the intended object and that results from separate tasks remain correctly associated.

Completed work or additional input

The specification distinguishes complete and input_required results. The latter indicates that an operation needs additional input before it can finish.

If a tool needs a user’s choice, the application can be evaluated on how it presents that step. Useful cases include answering, leaving the question unanswered and encountering an error during the request.

Another test can examine what users see when returning to the task. Is its status clear? Can they identify the object on which work will continue? These are design examples; assessing a particular application requires observing its actual behavior.

Extensions and product support

The MCP launch announcement also introduces a formal extensions framework and authorization updates. Anthropic announced a rollout of support across Claude products, describing adoption as a process being brought to its applications.

For an integration team, a practical starting point is to identify the client and server versions, then run a short task from beginning to result. The request, object identifier and outcome can be examined together, including a case that requires additional interaction.

Recording the expected behavior beforehand makes it easier to compare the actual result and locate a discrepancy.

The significance of MCP 2026-07-28 is its more explicit model for tool communication at the request level. Its implementation value becomes visible when both sides work together and the complete user journey can be followed and understood.

Sources

Sources: MCP maintainers — specification launch, 28 July 2026: https://blog.modelcontextprotocol.io/posts/2026-07-28/, MCP — versioned base protocol specification: https://modelcontextprotocol.io/specification/2026-07-28/basic, Anthropic — Claude rollout announcement, 28 July 2026: https://claude.com/resources/articles/bringing-mcp-2026-07-28-to-claude
Osamah Ismail

Founder and editor-in-chief of Kurd Developers. Phone and tech news on YouTube, Instagram and TikTok.