What's this about?
For Claude Team and Enterprise administrators, authorisation for MCP connectors is now centralised through the company's identity provider. Users get tools and data connected automatically, without each person clicking through an individual OAuth flow.
What exactly is MCP?
MCP is a standard way for an AI assistant to talk to external tools and data — your ticket system, your database, your file storage. Before a standard existed, every integration was bespoke. The security consequence is that each connection needs authorisation, and doing that per-user per-tool becomes unmanageable at company scale, which is exactly what this change addresses.
What's the community saying?
Discussion is mostly from administrators, and the recurring theme is that centralised authorisation is the precondition for deploying agents at all in a regulated environment — an individual OAuth prompt per employee per tool is not something a security team will approve.