It has been one year since Anthropic first open sourced Model Context Protocol (MCP), and to celebrate this anniversary a new version of the specification is being released.
“It’s hard to imagine that a little open-source experiment, a protocol to provide context to models, became the de-facto standard for this very scenario in less than twelve months,” the MCP Core Maintainers wrote in a blog post.
The latest release includes support for task-based workflows (experimental). According to the maintainers, tasks provide a new type of abstraction for tracking the work an MCP server performs. It enables several new capabilities, such as active polling to check the status of ongoing work anytime and result retrieval to see results of completed tasks. Tasks also support many different states including working, input_required, completed, failed, and cancelled.
Some use cases that tasks would be helpful for include healthcare data analysis where hundreds of thousands of data points are being processed, code migration tools, and multi-agent systems with agents working concurrently.
Another new update is support for URL-based client registration using OAuth Client ID Metadata Documents. This allows clients to provide a client ID as a URL pointing to a JSON document describing the client’s properties.
This is an alternative to some of the challenges associated with using Dynamic Client Registration (DCR) for authorization. To use DCR, MCP server developers need to have an Authorization Server (AS) that allows clients to register themselves through a public API, and if the AS doesn’t support that, the developer would need to build an OAuth proxy and manually register it with the AS and support DCR directly. A potential solution to this issue would be for every end user to provide their own client for registration, but the MCP maintainers say that would be “trading one complex task for another.” They believe this new URL-based client registration approach offers a more elegant solution to the problem.
The new version of the specification also includes security enhancements, such as new security requirements for clients used with local server installations, and an updated authorization specification that includes a default scopes definition.
Extensions—components that operate outside of the core specifications—are also being added. They allow developers to more easily build custom capabilities that follow MCP conventions without needing to fully integrate with the protocol.
“This approach allows for experimentation and specialized use cases while keeping the core protocol focused and stable. With extensions, we can move faster and enable developers to test out protocol capabilities before they become part of the specification,” the maintainers explained.
Building on this concept, the maintainers are also introducing authorization extensions, which allow additional authorization mechanisms outside of the specification to be implemented. The first two authorization extensions being released include support for OAuth client credentials for machine-to-machine authorization and enterprise IdP policy controls for MCP OAuth flows.
Other updates in this release of the MCP specification include:
- URL mode elicitation to allow users to be sent to a proper OAuth flow in their browser so they can authenticate without the client seeing their credentials
- The ability for MCP servers to run their own agentic loops using the client’s tokens
- A standardized format for tool names
- Decoupling request payloads from RPC methods definition
- SSE polling via server-side disconnect
- Improved specification version management for SDKs
The core maintainers also shared that their roadmap for MCP includes more work on reliability and observability, better patterns for server composition, and improvements to the security model.
“What excites us most isn’t what we’re planning to build but what our community is going to build. Every week we see MCP servers designed, developed, and deployed in novel ways. Every conversation in Discord reveals new use cases and patterns. The protocol has become a canvas for AI innovation, and we can’t fill it alone. The next year of MCP will be shaped by more production deployments, more real-world feedback, amplified by the creativity of thousands of developers worldwide. We’re here to support that growth, to ensure the protocol evolves thoughtfully, and to keep MCP stable, secure, and simple as it scales,” the maintainers wrote.
The post Several new updates make their way into the MCP specification appeared first on SD Times.
from SD Times https://ift.tt/nwfgGS0
Comments
Post a Comment