Intel / article

Vercel MCP now supports Client ID Metadata Documents

Source: vercel.com
Author: Josh Souphanthong
Date: Published Nov 2, 2026

Context

Vercel has opened its MCP server to any client that can identify itself in a newer way. , the Model Context Protocol, is a standard way for AI tools to connect to outside services. Until now, Vercel reviewed each client and its redirect URIs, the addresses a login flow sends the user back to, by hand, and rejected anything not on its . Users still log in and approve every connection, but a client no longer has to pass Vercel's review first.

The replacement is a Client ID Metadata Document, or CIMD: a JSON file the client publishes on its own domain to describe itself. The file's HTTPS address serves as the client's ID in OAuth, the standard behind the log-in-and-approve step. Vercel says this follows the MCP authorization specification, which now recommends CIMD over Dynamic Client Registration, a method in which each client registers itself with the server.

The trade-off appears on the approval screen. Because Vercel has not reviewed these clients, the screen shows where each one comes from: its name, its logo and the hostname that serves its metadata document, for example "Claude Code by claude.ai". CIMD clients do not get Vercel's verified badge, so that hostname is what a user is trusting when approving a connection. Apps already registered with Vercel work as before.

Key takeaways · AI-distilled

Vercel MCP previously accepted only clients that Vercel had reviewed and allowlisted by hand, along with their redirect URIs. Any MCP client can now request a connection, though the user still logs in and approves each one.
A client connects by publishing a Client ID Metadata Document, a JSON file on its own domain, and using that file's HTTPS URL as its OAuth client_id. Vercel notes the MCP authorization spec now recommends this over Dynamic Client Registration.
Because these clients are unreviewed, they get no Vercel verified badge. The consent screen shows the client's name, logo and the hostname serving its metadata document, so that hostname is what a user trusts when approving.

Terms in this piece · Glossary

MCP — The Model Context Protocol — an open standard that lets any AI assistant plug into any tool or data source without custom integration code.
tool permissions — The rules governing which tools an agent may call and which need confirmation — the boundary between a mistake and an incident.

Read the source vercel.com