What you can do with MCP Registry

Built for

MCP platform teams, directory operators, developer-tool builders, security reviewers, and agents that need current metadata about published MCP servers

Example workflows

Find registry entries

Turns a broad registry query into deployment-ready discovery context.

Try this

Search the MCP Registry for browser-related servers and summarize which entries have hosted remotes versus package-only installs.

Inspect a server

Grounds a server review in the current official registry metadata.

Try this

Fetch the latest registry entry for this MCP server name, then summarize its version, lifecycle status, repository, website, packages, and remotes.

Audit a namespace

Checks publication coverage and deployment mix for a registry namespace.

Try this

Analyze the `com.mcpbundles/` namespace in the MCP Registry and report active, deprecated, deleted, remote-capable, and package-only counts.

Track recent changes

Supports incremental directory sync and change monitoring.

Try this

List registry rows updated since this timestamp and group them by status and deployment type so we can decide what changed downstream.

Context to know first

Does the MCP Registry read API need an API key?

No. The public read endpoints are unauthenticated. Publishing and status edits use namespace-based authentication through the publisher flow.

What can agents inspect in the MCP Registry?

Agents can inspect server names, descriptions, versions, lifecycle status, packages, remote endpoints, repositories, websites, icons, and registry-managed timestamps.

Can agents search by description?

The official search parameter matches server names. Agents can scan pages and apply client-side filters, but the upstream API does not provide semantic description search.

AI Skill
SKILL.md

Domain knowledge for MCP Registry — workflow patterns, data models, and gotchas for your AI agent.

MCP Registry

The official MCP Registry is the public index of published MCP servers. Its read API is unauthenticated and exposes server metadata, version history, package entries, remote endpoints, repository links, website links, lifecycle status, and update timestamps.

Data Model

  • Server names use reverse-DNS form with one slash, such as io.github.owner/server or com.example/server.
  • Versions are immutable server.json snapshots. latest resolves to the current version for a server name.
  • Packages describe installable distributions such as npm, PyPI, NuGet, or OCI images plus transport details.
  • Remotes describe hosted MCP endpoints and their transport type.
  • Official metadata carries status, published time, update time, and whether the row is latest.

Workflows

  • Find servers — Search by server name substring, then inspect titles, descriptions, status, deployment types, and links.
  • Inspect one server — Fetch latest or a specific version before comparing packages, remotes, repository, and website metadata.
  • Track changes — Use RFC3339 updated_since filters for incremental registry monitoring; deleted rows can appear when incremental sync includes them.
  • Audit a namespace — Filter returned rows by namespace prefix such as com.mcpbundles/ to see publication coverage, statuses, and deployment surfaces.
  • Plan downstream ingestion — Prefer latest active rows with remotes when building a hosted-server directory; package-only rows can still matter for local stdio install flows.

Gotchas

  • Search is simple: the registry search parameter is a case-insensitive substring match on server names, not semantic search over descriptions.
  • Path encoding: server names contain /; detail and version requests must URL-encode the full server name.
  • Preview service: the registry is in preview and does not guarantee uptime or data durability. Treat it as a source to sync periodically, not as the only runtime data store.
  • Deleted status: deleted rows are hidden by default and mainly appear when explicitly requested or when using incremental sync.

Tools in this Server (5)

Mcp Registry Analyze Servers

Aggregate mirrored MCP Registry rows by status, namespace, deployment kind, package registry type, and remote transport type. Use namespace for publis...

Mcp Registry Compare Servers

Compare up to 20 mirrored MCP Registry server entries by full server name. Returns per-server status, version, package/remotes summary, links, and agg...

Mcp Registry Get Server

Fetch one MCP Registry server version from the local mirror by full server name. Use version=latest for the current mirrored row, or pass a specific v...

Mcp Registry List Versions

List all mirrored versions for one MCP Registry server name. The background refresh stores every version row it has seen from list pages.

Mcp Registry Search Servers

Search the local mirror of official MCP Registry server rows. Supports cursor pagination, namespace, lifecycle status, deployment type, and MCP App me...

Frequently Asked Questions

What is the MCP Registry MCP server?

Official Model Context Protocol Registry for discovering published MCP servers, their versions, deployment surfaces, and registry lifecycle status. It provides 5 tools that AI agents can use through the Model Context Protocol (MCP).

How do I connect MCP Registry to my AI agent?

Add the MCPBundles server URL to your MCP client configuration (Claude Desktop, Cursor, VS Code, etc.). The URL format is: https://mcp.mcpbundles.com/bundle/mcp-registry. Authentication is handled automatically.

How many tools does MCP Registry provide?

MCP Registry provides 5 tools that can be called by AI agents, along with a SKILL.md that gives your AI agent domain knowledge about when and how to use them.

What authentication does MCP Registry require?

MCP Registry uses open data APIs — no authentication required.

Does the MCP Registry read API need an API key?

No. The public read endpoints are unauthenticated. Publishing and status edits use namespace-based authentication through the publisher flow.

What can agents inspect in the MCP Registry?

Agents can inspect server names, descriptions, versions, lifecycle status, packages, remote endpoints, repositories, websites, icons, and registry-managed timestamps.

Can agents search by description?

The official search parameter matches server names. Agents can scan pages and apply client-side filters, but the upstream API does not provide semantic description search.

What is the difference between packages and remotes?

Packages describe installable server distributions, often for stdio use. Remotes describe hosted MCP endpoints that clients can connect to over a network transport.

Setup Instructions

Connect MCP Registry to any MCP client in minutes

https://mcp.mcpbundles.com/bundle/mcp-registry

One-click install:

The link prefills the Add custom connector dialog — you still review the values and click Add, then Connect to complete OAuth.

Or add manually

  1. Open claude.ai → Settings → Connectors.
  2. Click the + button and choose Add custom connector.
  3. Set Name to MCP Registry and paste the MCP URL into Remote MCP server URL.
  4. Click Add. MCP Registry will appear under Not connected — select it and click Connect to complete OAuth.
Name: MCP Registry
Remote MCP server URL: https://mcp.mcpbundles.com/bundle/mcp-registry
Authentication: OAuth

Custom connectors at claude.ai require a paid Claude plan (Pro, Max, Team, or Enterprise).

Try MCP Registry now

No API key or third-party login required. Chat with AI and run tools instantly.