Model Context Protocol adds enterprise authorization layer

The Model Context Protocol (MCP), an open standard introduced by Anthropic in November 2024 for connecting AI agents to tools, standardizes the wire format, roles and primitives but leaves memory, freshness and reliability to the implementing application. Adoption grew quickly, with 97 million monthly SDK downloads and 10,000 active servers at launch, and the protocol is governed by the Agentic AI Foundation with backing from Google, Microsoft, AWS, Cloudflare and Bloomberg.

A new "Enterprise‑Managed Authorization" extension now lets organizations control MCP server access centrally through their identity provider. The extension, first supported by Okta, uses the IETF‑draft Identity Assertion JWT Authorization Grant (ID‑JAG) to issue signed assertions that replace per‑tool OAuth prompts. Anthropic, Microsoft and other partners have integrated the feature, allowing agents such as Claude Managed Agents to be imported into corporate directories. "Logging in once and automatically having all your MCP connectors automatically set up is pretty magical," said Tom Moor, Head of Engineering at Linear. Okta’s Aaron Parecki described the change as making the identity provider "a centralized governance plane" for MCP access.