The Model Context Protocol (MCP) specification is maintained by the MCP Working Group, a collaborative community of developers and organizations focused on interoperability standards for machine learning systems. The technical documentation and version history are hosted on a public GitHub repository (e.g., github.com/mcpwg/spec), where contributors propose changes, review updates, and track issues. Governance typically follows a consensus-driven process, with major decisions requiring approval from a core maintainer team. This ensures the spec remains aligned with practical needs while avoiding fragmentation.
Maintenance involves regular reviews of proposed features, bug fixes, and community feedback. For example, the group might prioritize updates to support new model architectures (like transformer variants) or address edge cases in metadata formatting. The GitHub repository serves as the central hub for documentation, including the latest drafts, versioned releases, and implementation guides. Developers can submit pull requests or open discussions in the repository’s issue tracker to suggest improvements. Transparency in the process allows stakeholders to track progress and participate in shaping the spec.
The MCP spec is updated approximately every 3–6 months, though this varies based on community needs. Minor patches (e.g., v1.2.1) may roll out more frequently to fix critical bugs, while major releases (e.g., v2.0) occur after extensive testing and feedback cycles. For instance, the shift from MCP v1.0 to v2.0 in 2023 introduced support for dynamic model scaling, a feature requested by multiple framework maintainers. Release notes and upgrade guides accompany each version to help developers adopt changes smoothly. This balance of stability and adaptability ensures the spec stays relevant without disrupting existing implementations.