Model Context Protocol (MCP), one of the fastest-growing standards for connecting AI agents to business software and data, has received its biggest update since launching nearly two years ago.
The Linux Foundation has updated MCP to address several of the practical challenges that have limited large-scale enterprise deployments, including scalability, security, long-term compatibility and support for more sophisticated AI applications, according to the nonprofit’s blog post.
Originally developed by Anthropic and now governed by the Linux Foundation’s AI Alliance Foundation, MCP has become a common way for AI assistants and agents to connect with external applications, databases and software tools. Major AI companies, including OpenAI, Microsoft and Google, have adopted the protocol, helping establish it as a leading standard for agent interoperability.
Until now, however, many organizations have viewed MCP primarily as a tool for prototypes and early deployments rather than enterprise-wide production systems.
The latest update is designed to change that. It simplifies how MCP servers can be deployed across cloud infrastructure, making it easier for organizations to scale thousands of AI agent connections using existing cloud management tools. The release also strengthens authentication, introduces a formal deprecation policy that gives developers at least 12 months’ notice before features can be removed, and adds standardized support for long-running AI tasks and interactive applications.
Together, the changes are intended to make the protocol more stable and predictable for enterprises building AI systems expected to run continuously for years.
The update is significant because MCP is increasingly becoming the connective tissue between AI models and enterprise software. Rather than building custom integrations for every application, organizations can use the protocol as a common interface linking AI agents with business systems, internal data and external tools.