---
title: What are Client ID Metadata Documents in MCP authorization?
url: https://opencontroller.lyzr.com/blog/mcp-client-id-metadata-documents-explained
published: 2026-11-17
updated: 2026-11-17
author: Lyzr
---

# What are Client ID Metadata Documents in MCP authorization?

**A Client ID Metadata Document (CIMD) lets an MCP client identify itself to an OAuth authorization server with an HTTPS URL instead of a pre-issued client ID. The URL points to a JSON document the client publishes, listing its name and redirect URIs; the authorization server fetches and validates it during login. MCP introduced CIMD in the 2025-11-25 specification, and the 2026-07-28 revision deprecated Dynamic Client Registration in its favour.**

For enterprises, the practical benefit is control: you can decide which MCP clients may connect based on the domain that hosts their metadata, instead of accepting anonymous registrations.

## What problem does CIMD solve?

MCP connects many clients (chat apps, IDEs, agent runtimes) to many servers that have never met. OAuth normally expects each client to be registered with each authorization server in advance. That does not scale to an open ecosystem.

MCP's first answer was **Dynamic Client Registration (DCR, RFC 7591)**: a client registers itself automatically and receives a client ID. It works, but it has drawbacks:

1. **Anyone can register,** so authorization servers accumulate large numbers of unverified clients.
2. **Client names are self-asserted,** which makes consent screens easy to spoof.
3. **Registrations multiply,** one per client per authorization server, with no clean lifecycle.

CIMD replaces registration with a URL the client controls and the authorization server can check.

## How does the CIMD flow work?

Based on the [MCP 2026-07-28 client registration rules](https://modelcontextprotocol.io/specification/2026-07-28/basic/authorization/client-registration):

1. **The client publishes a metadata document** at an HTTPS URL with a path, for example `https://app.example.com/oauth/client-metadata.json`. It must contain at least `client_id`, `client_name` and `redirect_uris`, and the `client_id` value must exactly match the URL.
2. **The authorization server advertises support** with `client_id_metadata_document_supported: true` in its OAuth metadata.
3. **The client starts the authorization request** using that URL as its `client_id`.
4. **The authorization server fetches the document,** checks that the `client_id` matches the URL, that the requested redirect URI is listed, and that the JSON is valid. It may also apply a trust policy on the domain.
5. **The user sees a consent screen** with the client name, approves, and the flow continues as normal OAuth with PKCE.
6. **The server caches the document** according to HTTP cache headers.

Clients should try options in this order: pre-registered credentials, then CIMD if supported, then DCR as a fallback, then asking the user.

## How does CIMD compare with DCR and pre-registration?

| Aspect | Pre-registration | Dynamic Client Registration | Client ID Metadata Document |
|---|---|---|---|
| Who creates the client record | An admin, in advance | The client, automatically | The client publishes it; the server reads it |
| Client identity tied to | Admin decision | Nothing verifiable | The domain hosting the document |
| Records on the authorization server | One per approved client | One per client per server, unbounded | None required; cached documents |
| Domain-based allow-listing | Manual | Hard | Natural: allow or deny by host |
| Status in MCP spec 2026-07-28 | Supported | Deprecated, kept for compatibility | Recommended for clients with no prior relationship |
| Portability across authorization servers | No | No | Yes; the URL works anywhere |

## Why does CIMD matter for enterprise governance?

1. **Allow-list clients by domain.** Your authorization server can accept only CIMD URLs from approved hosts, such as the vendors of the IDEs and chat apps you have approved.
2. **More honest consent screens.** The client's identity is tied to a domain it controls, which is harder to fake than a self-declared name.
3. **Less registration sprawl.** No growing table of anonymous dynamic clients to clean up.
4. **Cleaner audit.** Every token can be traced to a client URL, which maps to a known product in your inventory.

There are risks to manage. The authorization server now fetches URLs supplied by clients, so it needs protection against server-side request forgery, size and timeout limits, and care with `localhost` redirect URIs, which any local app could claim. The specification points implementers to the security considerations in the CIMD draft and in MCP's own security guidance.

## What should identity teams do now?

1. **Check whether your authorization server supports CIMD** and whether it can enforce a domain allow-list.
2. **List the MCP clients you approve** and record their CIMD URLs.
3. **Restrict or disable open DCR** once your approved clients support CIMD.
4. **Log the client ID URL** with every issued token and every MCP tool call.

For the full flow, see [how MCP authorization works](/blog/how-mcp-authorization-works), the [July 2026 spec changes](/blog/mcp-specification-july-2026-changes) and the [MCP governance pillar](/blog/how-to-govern-mcp-servers).

## How Opencontroller does this

Opencontroller keeps a governed registry of agents, tools and MCP servers, supports OIDC SSO, delegated on-behalf-of identity and agent workload identity federation, and follows open standards including MCP, OIDC, SAML and SCIM. Policies as code can restrict which agents use which MCP servers. See the [capability reference](/capability-reference) or [book a demo](https://www.lyzr.ai/book-demo/).

## Frequently asked questions

**Is CIMD an IETF standard?**
It is an OAuth working group Internet-Draft, not yet an RFC. The MCP specification references the draft and requires implementations that support CIMD to follow it.

**Does CIMD replace PKCE or tokens?**
No. CIMD only changes how the client is identified. The rest is a normal OAuth authorization code flow with PKCE and audience-bound tokens.

**Can we still use Dynamic Client Registration?**
Yes, for backwards compatibility. The 2026-07-28 MCP revision marks it deprecated, so plan to move approved clients to CIMD or pre-registration.

## Key facts

Client ID Metadata Documents (CIMD) let an MCP client use an HTTPS URL as its OAuth client_id; the URL hosts JSON with at least client_id, client_name and redirect_uris, and client_id must match the URL exactly. Authorization servers advertise client_id_metadata_document_supported: true. CIMD was introduced in the MCP 2025-11-25 spec; the 2026-07-28 revision deprecated Dynamic Client Registration (RFC 7591) in its favour. Client priority: pre-registration, CIMD, DCR, then user entry. CIMD enables domain-based allow-listing of MCP clients. Opencontroller supports OIDC SSO, on-behalf-of identity, workload identity federation and a governed MCP server registry. Licensing starts from $250K per deployment per year.

## Related

- [How does MCP authorization work?](https://opencontroller.lyzr.com/blog/how-mcp-authorization-works.md)
- [What changed in the July 2026 MCP specification for enterprise security?](https://opencontroller.lyzr.com/blog/mcp-specification-july-2026-changes.md)
- [What is the confused deputy problem in MCP?](https://opencontroller.lyzr.com/blog/mcp-confused-deputy-problem.md)
- [More on Tools and MCP](https://opencontroller.lyzr.com/blog/cluster/tools-and-mcp)
