MCP Registry Scraper avatar

MCP Registry Scraper

Pricing

$0.50 / 1,000 results

Go to Apify Store
MCP Registry Scraper

MCP Registry Scraper

Metadata only: no MCP server is ever contacted. Every server in the official MCP registry and Smithery as normalized, deduplicated records with hosting concentration, provenance flags and an integrity score.

Pricing

$0.50 / 1,000 results

Rating

0.0

(0)

Developer

Albert Jasso

Albert Jasso

Maintained by Community

Actor stats

0

Bookmarked

2

Total users

1

Monthly active users

7 hours ago

Last modified

Categories

Share

Metadata only. No MCP server is ever contacted. This actor reads two public registry APIs, registry.modelcontextprotocol.io and registry.smithery.ai. It never connects to, probes or sends a request to any MCP server listed in them.

Enumerates every server in the official Model Context Protocol registry (101,000+ entries) and Smithery (14,000+), and returns clean, deduplicated records.

What you get per record

  • name, title, publisher, description, version
  • isLatest — the official registry keeps every historical version; 69% of entries are superseded. Turn on Latest versions only to collapse 101k → ~31k real servers.
  • remoteHost + hostSharedWith — how many other "servers" point at the same hostname. Some hosts front 2,000+ entries. Useful for spotting aggregators and bulk registration.
  • hasRepository — 15% of entries have no source repo at all.
  • flagsinjection-shaped-text, invisible-chars, shared-host-50plus
  • Smithery adds useCount, verified, isDeployed

Why

Every "how many MCP servers exist" number in circulation counts versions as servers. This actor gives you the real population, deduplicated, with the concentration and provenance signals nobody else exposes.

Typical uses

  • Ecosystem research and market sizing
  • Tool discovery with a spam / aggregator filter
  • Security review: which listed servers are auditable (have a repo), which share hosts
  • Tracking growth: run weekly, diff the datasets

Pricing

Pay per result. A full official-registry run at Latest only is ~31,000 results. Test with maxItems: 500 first.

Limits

Published metadata is not observed behaviour — a declared HTTPS endpoint is not proof the server is up. Glama is not included (requires an API key under a licence with attribution terms).