MCP Registry Preview: A Federated DNS for Enterprise AI Context
What the MCP Registry is
The MCP team has released a preview of the MCP Registry, a federated discovery layer designed to make enterprise AI context discoverable and governable at scale. More than a simple catalog, the Registry provides a standardized mechanism for registering and locating MCP servers, whether public or private, and models addressability in a way similar to how DNS solved internet addressing.
Registry as DNS for AI context
At its core, the MCP Registry serves as DNS for AI context. Organizations such as GitHub or Atlassian can publish public MCP servers to a shared directory while enterprises can run private sub-registries. This two-tier approach creates a secure front door into the broader MCP ecosystem without exposing internal infrastructure.
A single global registry would introduce significant security and compliance risks. The federated design provides an authoritative upstream source of truth while allowing organizations to apply their own access controls, policies, and operational rules locally.
Why a federated model matters
Enterprises commonly operate in hybrid environments that combine internal data and external services. The Registry acknowledges that reality and enables a range of practical use cases:
- Secure internal discovery of context servers without exposing private endpoints to the public internet
- Centralized governance and audit trails to control which external MCP servers are allowed
- Reduced context sprawl by encouraging teams to adopt a single protocol and discovery layer
- Support for hybrid AI agents that can query private MCP servers alongside public documentation sources within the same framework
By distributing responsibility across sub-registries while maintaining a shared upstream reference, the Registry balances scalability with security and compliance.
Architecture, moderation, and open source foundation
The MCP Registry is an open project under a permissive license and is currently available in preview. It provides an upstream API specification that sub-registries can inherit to ensure interoperability. Public marketplaces can extend upstream data for specific needs and private registries can enforce organizational policies, moderation rules, and governance requirements.
The preview is managed by the MCP registry working group and aims to foster interoperable implementations while allowing enterprises to retain control over private context.
Getting started
The Registry is available in preview. To begin using it, follow the MCP guides for adding servers and accessing registry data. The preview label means features may change and durability guarantees are not yet provided until general availability.
FAQs
- What is the MCP Registry? It is a global directory and API for discovering MCP servers, acting like DNS for AI context and supporting both public catalogs and enterprise sub-registries.
- Why a federated model? A single registry would pose compliance and security risks. Federation allows private sub-registries to coexist with a shared upstream source of truth.
- How do enterprises benefit? Enterprises gain secure internal discovery, centralized governance, prevention of context sprawl, and support for hybrid AI agents.
- Is it open source? Yes. The Registry is an official MCP project with permissive licensing, API specs, and tools for sub-registry development.
- Is it GA? Not yet. The Registry is in preview and may change before general availability.