registry/docs/reference/server-json/official-registry-requirements.md
Radoslav Dimitrov 7a6d1bc30a
fix(cargo): harden README fetch, clarify status handling, strengthen tests (#1330)
Follow-up to #1207, addressing items from the cargo validator review.
**Scoped to cargo + shared helpers — no behavior change for existing
npm/pypi/nuget/oci/mcpb publishers.** Safe to merge and promote.

### Changes
**SSRF hardening of the two-call README retrieval**
- Pin the step-2 README URL to an allowed host (`static.crates.io` in
prod; the base host under test) *before* fetching, so a metadata
response can't steer the validator at an internal/attacker host.
- `CheckRedirect` policy pins every redirect hop to the same allowlist
(the initial URL being pinned isn't enough if an upstream 3xx is
followed).
- Cap the README body with `io.LimitReader` (5 MiB).

> Not publisher-exploitable today (the URL comes from crates.io), so
this is defense-in-depth — but it makes the "host is pinned" guarantee
hold in code rather than rely on crates.io's behavior.

**Clearer status handling** (previously any non-5xx/non-200 collapsed to
"not found")
- **429** → reported as transient/retryable, at both the metadata and
README steps.
- **403** → disambiguated via the crate-version endpoint:
genuinely-missing stays "not found"; *exists-but-no-README* gets an
actionable "add a README with `mcp-name` and republish" message (mirrors
the NuGet validator). This continues the 5xx-vs-403 diagnostic split
@P4ST4S started in the #1207 review — thanks!

**Shared `containsMCPNameToken` helper**
Boundary-anchored ownership-token match (prevents prefix confusion, e.g.
a README declaring `io.github.acme/widget-pro` satisfying a claim for
`io.github.acme/widget`), used here **by the cargo validator only**.
Adopting it in PyPI/NuGet is a separate follow-up because it's a
behavior change for those already-live registries.

**Tests & docs**
Hermetic positive `ServerNameFormats`; combined fixture gains a
version-existence endpoint + cases for 403-missing vs 403-no-README,
429, and prefix-confusion rejection; new foreign-README-host (SSRF)
test; internal test for the helper. Docs: list `crates.io` in the
registry-requirements list + the `Package` model doc.

### Testing
`go build`, `go vet`, full `./internal/validators/...` suite (incl. live
cargo positive), `make check-schema`, `validate-examples` all pass. No
new `golangci-lint` findings.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-06-04 17:34:54 -04:00

5.8 KiB

Official Registry Server.json Requirements

This document describes the additional requirements and validation rules that apply when publishing to the official MCP Registry at registry.modelcontextprotocol.io.

For step-by-step publishing instructions, see the publishing guide.

Overview

While the generic server.json format defines the base specification, the official registry enforces additional validation to ensure:

  • Namespace authentication - Servers are published under appropriate namespaces
  • Package ownership verification - Publishers actually control referenced packages
  • Restricted registry base urls - Packages are from trusted public registries
  • _meta namespace restrictions - Restricted to publisher key only

Namespace Authentication

Publishers must prove ownership of their namespace. For example to publish to com.example/server, the publisher must prove they own the example.com domain.

See the publishing guide for authentication details for GitHub and domain namespaces.

Package Ownership Verification

All packages must include metadata proving the publisher owns them. This prevents impersonation and ensures authenticity (see more reasoning in #96).

For detailed verification requirements for each registry type, see the publishing guide.

Restricted Registry Base URLs

Only trusted public registries are supported. Private registries and alternative mirrors are not allowed.

Supported registries:

  • NPM: https://registry.npmjs.org only
  • PyPI: https://pypi.org only
  • NuGet: https://api.nuget.org/v3/index.json only
  • Cargo: https://crates.io only
  • Docker/OCI:
    • Docker Hub (docker.io)
    • GitHub Container Registry (ghcr.io)
    • Quay.io (quay.io)
    • Google Artifact Registry (*.pkg.dev)
    • Azure Container Registry (*.azurecr.io)
    • Microsoft Container Registry (mcr.microsoft.com)
  • MCPB: https://github.com releases and https://gitlab.com releases only

_meta Namespace Restrictions

The _meta field in server.json allows publishers to include custom metadata alongside their server definitions.

Publisher-Provided Metadata

When publishing to the official registry, only data under the specific key io.modelcontextprotocol.registry/publisher-provided will be preserved. Any other keys in the _meta object are silently dropped during publishing and will not be stored or returned by the registry.

Example:

{
  "_meta": {
    "io.modelcontextprotocol.registry/publisher-provided": {
      "tool": "ci-publisher",
      "version": "1.0.0",
      "custom_data": "your data here"
    },
    "some.other.key": {
      // This will be dropped and not preserved
    }
  }
}

Size limit: The publisher-provided extension is limited to 4KB (4096 bytes) of JSON. If the marshaled JSON exceeds this limit, publishing will fail with an error indicating the actual size.

When your publisher-provided metadata gets large or comes from multiple sources (e.g. GitHub-specific hints alongside your own CI tool's metadata), grouping keys under reverse-DNS subkeys is a useful convention to avoid accidental collisions:

{
  "_meta": {
    "io.modelcontextprotocol.registry/publisher-provided": {
      "com.github": {
        "serverDisplayName": "My Server"
      },
      "io.example.ci-publisher": {
        "tool": "ci-publisher",
        "version": "1.0.0"
      }
    }
  }
}

This is not enforced; flat keys work fine for simple cases, and existing examples in this repo use the flat form. Use the namespaced form when organizing metadata from more than one source.

Registry API Metadata vs server.json Metadata

The _meta field in server.json is different from the _meta field returned in registry API responses:

  • In server.json: The _meta field contains publisher-provided custom metadata under io.modelcontextprotocol.registry/publisher-provided
  • In API responses: The _meta field is returned as a separate property at the response level (not inside server.json) and contains registry-managed metadata like:
    • status: Server lifecycle status (active, deprecated, deleted)
    • publishedAt: When the server was first published
    • updatedAt: When the server was last updated
    • isLatest: Whether this is the latest version

Example: What you publish (server.json)

{
  "name": "io.github.example/my-server",
  "version": "1.0.0",
  "description": "My MCP server",
  // ... other fields ...
  "_meta": {
    "io.modelcontextprotocol.registry/publisher-provided": {
      "tool": "ci-publisher",
      "version": "2.0.0"
    }
  }
}

Example: What the registry API returns

{
  "server": {
    "name": "io.github.example/my-server",
    "version": "1.0.0",
    "description": "My MCP server",
    // ... other fields ...
    "_meta": {
      "io.modelcontextprotocol.registry/publisher-provided": {
        "tool": "ci-publisher",
        "version": "2.0.0"
      }
    }
  },
  "_meta": {
    // Registry-managed metadata at response level
    "status": "active",
    "publishedAt": "2024-01-15T10:30:00Z",
    "updatedAt": "2024-01-15T10:30:00Z",
    "isLatest": true
  }
}

Notice how the registry API response has two _meta fields:

  1. Inside the server object: Your publisher-provided metadata (preserved from server.json)
  2. At the response level: Registry-managed metadata (automatically added by the registry)

Registry-managed metadata cannot be set or overridden by publishers. See the API documentation for the complete response structure.