MCP Registry Indexer + Capability Matcher

The only Apify actor built on the canonical, protocol-maintained Model Context Protocol registry — registry.modelcontextprotocol.io — instead of a third-party MCP directory. It turns the living MCP server catalog into clean, RAG-ready records and answers the question every AI-agent builder actually has: "which MCP server does my agent need for this task?"

What it does

1. Indexes the full canonical registry into normalized records: server name, title, description, transport (streamable-http / stdio / sse), remote URL, protocol version, auth requirement, and derived capability tags.
2. Matches a free-text task (taskQuery) to servers using intent-aware synonym expansion (e.g. "search the web and write to a Google Sheet" expands to web/search/google/sheet tokens) and per-field scoring, so the strongest title/tag hits rank first.
3. Monitors freshness (optional) — stores a snapshot in the actor KV store and flags servers added or removed since the previous run, so you can detect registry drift or deprecated servers.

Input example

{
  "taskQuery": "search the web and write the result into a Google Sheet",
  "minMatchScore": 20,
  "includeCapabilities": true,
  "freshnessMonitor": true,
  "maxServers": 300
}

Leave taskQuery empty to index the whole registry without matching. maxServers: 0 fetches every server (needs a long run timeout — the registry is large).

Output example (match record)

{
  "kind": "match",
  "server_name": "ai.exa/exa",
  "server_title": "Exa",
  "description": "Search the web with embeddings. The Exa MCP server exposes web search as a tool.",
  "transport": "streamable-http",
  "remote_url": "https://mcp.exa.ai/mcp",
  "protocol_version": "1.0.0",
  "auth": "unspecified",
  "tags": ["web", "search", "api"],
  "capabilities": ["web", "search", "api"],
  "tools": [],
  "match_score": 60,
  "matched_on": ["search", "web"],
  "registry_url": "https://registry.modelcontextprotocol.io/ai.exa/exa"
}

A real run against the live registry (capped at 300 servers) returned 282 unique servers and 20 task matches. Top hits for the query above: Exa web search (60), Google Search Console (60), Google News (40), Web Analytics (30). Index records carry kind: "server" with the same fields minus match_score/matched_on.

How to use

1. Open the Actor and paste the JSON input (use the Input tab's JSON editor, or the form fields).
2. Set taskQuery to a plain-English task, or leave it empty to just index.
3. Turn on freshnessMonitor if you want drift detection between scheduled runs.
4. Click Start. Records appear in the default dataset — export to JSON, CSV, or pull via API.

Source notes

The registry returns one entry per version, so a page of 100 entries holds ~68 unique servers — the Actor dedupes by server.name, keeping the entry flagged isLatest. Pagination is cursor-based via metadata.nextCursor passed as ?cursor= (?offset= is ignored). Capabilities are derived from the server description (the registry does not expose a tools array); auth is reported as unspecified because the registry does not publish per-server auth requirements.

Pricing

Pay-per-event (PPE): a small per-run start fee plus a per-record item fee. The source is public, keyless JSON — no proxy, no auth, no per-call API cost. See the Actor's pricing tab for the current rate.

Compliance

Reads only publicly published registry metadata (server name, description, transport, tool list). No personal data, no login, no platform ToS conflict. For production agent connectivity, prefer the official registry over scraping any single vendor's site.

Support

Found a bug or need a field added? Open an issue on the Actor's Issues tab.