# Azure API Center Azure API Center is Microsoft Azure's solution for design-time API governance and centralized discovery. It keeps a structured inventory of all of an organization's APIs (any type, lifecycle stage, or deployment location) along with versions, definition files, and metadata, so program managers and developers can discover, reuse, and govern them. It is the design-time complement to Azure API Management, which handles runtime governance through a gateway. ## Key capabilities - **Inventory**: register every API (managed, unmanaged, or in development) from the Azure portal, Azure CLI, CI/CD, or links to sources like API Management - **Governance**: custom metadata, plus linting and analysis to enforce API style guidelines and catch breaking changes (shift-left in VS Code) - **Discovery**: an Azure API Center portal where developers browse and filter the inventory by built-in and custom metadata - **Shadow-API detection**: integrates with Dev Proxy so apps don't call unregistered APIs ## Catalog for MCP servers, AI Skills, and AI Plugins Beyond classic REST APIs, API Center can be used to build and maintain an enterprise catalog of AI assets: [[Model Context Protocol (MCP)|MCP]] servers, [[AI Agent Skills|AI Skills]], AI Plugins, and more. The catalog respects the [[Model Context Protocol Registry]] specification, so it interoperates with the broader MCP ecosystem rather than being a closed silo. An example portal is `mcp.azure.com`. A catalog like this is not only for discovery. It is also an enforcement point: an org can curate the approved set of MCP servers / skills and require clients to use only what is listed (allowlisting), giving central governance over which AI capabilities employees can wire into their agents. ### As a private MCP registry API Center exposes an MCP registry endpoint implementing the registry v0.1 spec, at `https://<name>.data.<region>.azure-apicenter.ms/workspaces/default/v0.1/servers` (the `workspaces/default` segment is mandatory). Per MCP server, the inventory holds the discovery and configuration metadata clients need: remote endpoints with transport type (SSE / streamable HTTP) and environment, or, for local servers, the package registry, package name, version, runtime hint (e.g. `npx`), and runtime arguments. Custom metadata can be mapped into the `_meta` section of MCP responses. A curated list of partner MCP servers (Microsoft services, GitHub, Atlassian, ...) can be registered in one click, and the inventory can auto-sync from Azure API Management or a Git repository. Registered servers also feed Microsoft Foundry's tool catalogs for AI agents. For remote MCP servers, API Center auto-generates OpenAPI definitions for both SSE and streamable HTTP transports. The API Center portal gives stakeholders a browsable catalog with per-server tool schemas, one-click install into [[Visual Studio Code (VSCode)]], and a built-in test console (MCP Inspector) to run tools and see responses without leaving the portal. Access management controls who can view and use which servers. The main consumer: [[GitHub Copilot MCP Registry]]. Point a GitHub org/enterprise's Copilot policy at the registry URL and approved servers become discoverable (and, with "Registry only", the enforced allowlist) across Copilot clients: `@MCP` in [[Visual Studio Code (VSCode)]] and `/mcp search` in [[GitHub Copilot CLI]] then list only what the registry serves. Caveat: the endpoint currently needs anonymous read access, so no secrets in server metadata. ## Plans Free plan (evaluation, small scale) and Standard plan (enterprise). Linking an eligible API Management instance grants the Standard plan at no extra cost. ## References - https://learn.microsoft.com/en-us/azure/api-center/overview - https://learn.microsoft.com/en-us/azure/api-center/ - https://github.com/Azure/api-center - https://mcp.azure.com/ - https://www.youtube.com/watch?v=7h0UxwYHXc0 - MCP server inventory: https://learn.microsoft.com/en-us/azure/api-center/register-discover-mcp-server - Enforcement walkthrough: https://devblogs.microsoft.com/all-things-azure/locking-down-mcp-create-a-private-registry-on-azure-api-center-and-enforce-it-in-github-copilot-and-vs-code/ ## Related - [[Model Context Protocol (MCP)]] - [[Model Context Protocol Registry]] - [[AI Agent Skills]] - [[Docker MCP Catalog]] - [[MCP Gateway Registry]] - [[AI Agents]] - [[GitHub Copilot MCP Registry]] - [[GitHub Copilot]] - [[GitHub Copilot CLI]]