Discover and manage MCP servers to avoid duplication and complexity

Companies use the complementary Agent2Agent (A2A) and Model Context Protocol (MCP) standards to build robust AI agents. The former provides agent-to-agent communication and discoverability, while the latter enables agent-to-tool communication, standardizing how AI agents interact with tools, APIs, and resources to obtain information.

We released the discovery of MCP servers registered in the Microsoft Azure API Center. The discovery process identifies your MCP servers along with metadata such as name, description, URL, and lifecycle stage, and lists them in the AI agent discovery inbox. There, you can review them and decide to move them into your inventory. Once moved into your inventory, they will be modeled as fact sheets of the interface subtype "MCP server". By linking MCP servers to applications and AI agents, you gain a holistic view of connections and dependencies of AI-related components in your enterprise architecture landscape. This approach helps identify additional optimization opportunities, such as building agent chains, and promotes the reuse of existing tools.

Over the course of the upcoming quarters, we plan to include additional MCP server information in our reference catalog to further support the discovery and management of MCP servers.