MCP Python SDK v2.1.0 Boosts Features, Claude Code v2.1.245 Stabilizes Linux

The Model Context Protocol (MCP) Python SDK has advanced to version 2.1.0, delivering key enhancements for developers working with AI models and multimodal interactions. A notable highlight of this release is the expanded capability for prompt messages, which now accept `Image` and `Audio` data types. This means that AI applications built with the SDK can integrate richer, more diverse inputs beyond just text, facilitating more sophisticated multimodal experiences. Furthermore, prompt functions within the protocol may now return bare content, simplifying the development of tools and agents that interact with models. Another significant improvement is the streamlining of client setup: the `Client` object now directly accepts `StdioServerParameters`. This allows for a more concise and intuitive configuration of server commands, exemplified by `Client(StdioServerParameters(command="uv", args=["run", "server.py"]))`. This change reduces boilerplate code and makes it easier for developers to deploy and manage AI services, ensuring greater flexibility and efficiency in building MCP-compliant applications. The update reinforces MCP's commitment to supporting complex AI workflows with practical, developer-friendly tools.

Anthropic has released Claude Code v2.1.245, a targeted maintenance update designed to significantly improve the stability of the application on various Linux distributions. This specific release primarily addresses a critical issue that caused a crash on startup for Linux systems utilizing `glibc 2.44`. The bug impacted several popular and actively maintained distributions, including Arch Linux, CachyOS, and Fedora Rawhide. By providing a fix for this specific `glibc` compatibility problem, Anthropic ensures that developers and researchers running Claude Code in these environments can now launch and operate the application without encountering unexpected termination. This enhancement is vital for maintaining productivity and reducing friction for a significant portion of the user base that relies on these modern Linux platforms. The update focuses solely on resolving this critical compatibility issue, reflecting a commitment to reliability and a stable development experience for all users of Claude Code.

Google has published a new guide for developers focusing on the crucial task of evaluating live and voice AI agents within the ADK framework, a necessary step for successfully transitioning agents from initial demonstration phases to robust production deployments. The article underscores the critical need for rigorous, automated testing methodologies to effectively manage the inherent unpredictability and complexity of real-world, multi-turn conversations. To facilitate this, the ADK now provides native live evaluation capabilities, empowering developers to systematically test and refine agent performance under conditions that closely mimic actual user interactions. This feature allows for the simulation of complex conversational scenarios, which is invaluable for identifying and mitigating potential issues related to dialogue flow, intent recognition, and response generation before an agent goes live. The guide also offers practical insights into best practices for setting up these comprehensive evaluations, with a strong emphasis on establishing metrics that accurately reflect an agent's ability to handle diverse user inputs, manage context, and maintain conversational coherence across extended interactions. This robust testing framework is designed to ensure that agents can reliably perform in demanding live settings, adeptly handling the nuances and challenges of authentic human interaction.