MCP Team Launches Preview of “MCP Registry”: Joint Discovery Layer for Enterprise AI

Model Context Protocol (MCP) team has released MCP Registrationthe system could be the ultimate puzzle that makes enterprise AI really ready for production. The MCP registry is not just a directory, it introduces a federation architecture for discovering MCP servers (public or private) that reflects how the Internet itself resolves addressability decades ago.

DNS with registry for AI context

The MCP registry is centered on this DNS for AI context. It provides a global public directory where companies such as Github or Atlassian can publish MCP servers, while also providing businesses with a standardized way to operate private subregistration. This two-layer approach creates a secure “front door” for the wider MCP ecosystem without compromising internal privacy.

A single registry will create untenable security and compliance risks. In contrast, the federal model hits the need to balance firms: the authoritative upstream sources of truth, and the flexibility to extend or limit it using organization-specific rules.

Why the joint model works?

Businesses run in a hybrid environment – foster internal systems and external services. The registry design recognizes that it realistically implements use cases that were not easy before:

• Security internal discovery: Teams can discover and consume internal servers (such as “customer support context”) without exposing private infrastructure to the Internet.
• Centralized governance: Which external MCP servers can enterprises use to access which external MCP servers and have a complete audit trail to comply with regulations.
• Reduce context spread: Rather than custom temporary integration, it aligns around a team in a single protocol and governance layer.
• Hybrid AI Agent: Agents can seamlessly query private data (via the internal MCP server) and public documents (via GitHub’s MCP server) within the same framework.

The result is a regulated, scalable infrastructure layer that unifies AI proxy connections across boundaries.

Architecture, temperance and open source foundation

The MCP registry is Open Project With a spacious license, now available in preview, managed by the MCP Registry Working Group. It provides an upstream API specification that can be inherited to ensure interoperability. Public “markets” can enhance upstream data to meet specific customer needs, while private enterprise registration agencies can enforce internal policies.

Summary

For enterprises, a stable version of the MCP registry can provide missing connective tissue between a private environment and a public AI infrastructure. It can eliminate the breakdown and risk of uncontrolled integration through standardized discovery and governance. The architecture can be firmly scaled – because it assigns responsibility while maintaining the upstream source of truth.

MCP registration performance is available in preview. start:

FAQ

FAQ 1: What is the MCP Registry?
The MCP Registry is a global directory and API for discovering MCP servers. Its role is similar to DNS in the AI ​​context, allowing both public directories and enterprise subregistration agencies to perform.

FAQ 2: Why does the registry use a federated model instead of a single global registry?
A single registry poses compliance and security risks. The joint model allows businesses to run private subregistration activities while relying on sharing sources of upstream truth.

FAQ 3: How do businesses benefit from the MCP registry?
Enterprises gain secure internal discovery, centralized governance of external servers, preventing context spreads, and support for hybrid AI agents.

FAQ 4: Is the MCP registry open source?
Yes. It is an official MCP project, open source and licensed, with its API and specifications available for subregistration development.

FAQ 5: Is the MCP registry usually available?
not yet. The MCP registry is currently in preview mode, which means that features may change and no durability is provided until general availability.