Industry Giants Back New Agent Plugin Spec, but Anthropic Remains on the Sidelines
Vercel's new portability standard aims to end AI fragmentation, yet the author of its underlying logic is missing from governance.
The AI industry has taken a significant step toward interoperability with the release of Agent Plugins 1.0.0, a specification designed to make AI agent components portable across different platforms. While the move signals a rare moment of alignment among tech giants, a notable gap in governance has emerged regarding the company that provided the technical foundation for the project.
Vercel released the Agent Plugins 1.0.0 specification with formal support from a coalition of core maintainers including AWS, Cursor, Microsoft, OpenAI, and Google. The specification focuses on the logistics of portability, specifically defining the packaging and discovery of components—essentially establishing where plugins live and how they are found. However, the spec does not standardize the actual behavior or skill formats of these agents. Instead, it defers those requirements to the Agent Skills specification, which mandates a SKILL.md file containing specific YAML frontmatter—including a name and description—followed by a Markdown body.
The Fragmentation Problem
This push for standardization comes as developers struggle with extreme fragmentation in the agentic AI ecosystem. Currently, a single tool or skill often must be rewritten or forked to function across different agent clients, such as moving a capability from a custom Claude-based agent to one powered by OpenAI. This lack of portability creates high maintenance overhead and slows the deployment of complex, multi-platform AI workflows.
The Governance Gap
Despite the broad industry support for the packaging spec, questions have emerged regarding the role of Anthropic. While Anthropic is not listed as a core maintainer of the Agent Plugins governance, the company authored the underlying Model Context Protocol (MCP) and Agent Skills standards that the Agent Plugins spec bundles. This creates a strange paradox: the industry is adopting a portability format built on Anthropic's logic, yet Anthropic is not actively helping govern the broader format.
Why It Matters
Without a unified governance body that covers both the packaging of plugins and their actual behavior, the industry risks a partial solution. The absence of a key architectural player like Anthropic from the formal governance of the Agent Plugins spec suggests a potential split in how the industry approaches interoperability. If the packaging standard and the behavioral standards diverge in their leadership, developers may still face friction when attempting to scale agents across different ecosystems.
What's Next
Industry observers are now watching to see if Anthropic will formally join the governance coalition or if it will continue to push its own standards independently via MCP. Until a comprehensive standard is managed by a truly representative body, the promise of "write once, run anywhere" for AI agents remains an aspiration rather than a reality.