MCP Server Catalog + Quality Score

mcp-server-catalog is an Apify Actor that scrapes the top awesome-mcp-server GitHub lists, scores every MCP (Model Context Protocol) server on six quality dimensions (stars, recency, license, description, docs, activity), and returns a ranked dataset of production-ready MCP servers. Use it to skip clicking through 500+ READMEs and instantly find which MCP server is worth installing.

Best for / Not for

✓ Best for	✗ Not for
Pre-install due diligence before adding an MCP server to Claude / Cursor / Cline / Aider	Real-time monitoring of one specific MCP server (use a different tool)
Security teams looking for unlicensed or abandoned MCP servers in their supply chain	Servers not listed in any awesome-mcp catalog (this is a list-scraper)
Investors / analysts tracking MCP ecosystem growth over time via scheduled runs	Live runtime metrics — this is metadata + GitHub stats, not behavior

Example input → output

Input:

{
  "keywordFilter": "github",
  "maxServers": 100,
  "sourceLists": ["punkpeye/awesome-mcp-servers"]
}

Output (one row, truncated):

{
  "fullName": "github/github-mcp-server",
  "name": "github-mcp-server",
  "url": "https://github.com/github/github-mcp-server",
  "sourceList": "punkpeye/awesome-mcp-servers",
  "stars": 12500,
  "forks": 890,
  "openIssues": 45,
  "lastCommit": "2026-05-18T14:23:11Z",
  "license": "MIT",
  "language": "TypeScript",
  "description": "Official MCP server for GitHub API",
  "qualityScore": 95,
  "scoreBreakdown": {
    "stars": 30, "recency": 20, "license": 15,
    "description": 10, "docs": 10, "activity": 15
  }
}

A Markdown report (mcp-catalog.md) is also written for human-readable browsing.

FAQ

Does this work with Claude Code?

Yes — the output dataset and ranked Markdown report are exactly what you need before adding any MCP server to your Claude Code .mcp.json or settings. Pipe the JSON into a claude mcp add script to bulk-install only the highest-scoring ones.

Is it free?

The first run is a free trial (up to 50 servers scored). After that the Actor uses Apify's pay-per-event model at $0.02 per server scored ($2 for a 100-server catalog). A $5/month subscription gives you a daily-refreshed catalog.

Does it scan private repos?

No — it reads public GitHub data only. Private MCP servers won't appear in the awesome-mcp lists this Actor scrapes, so they wouldn't be in scope anyway.

Can I run it in CI?

Yes. Schedule it via Apify cron (weekly recommended) and diff the dataset between runs to catch new high-quality MCP servers as they appear. Or trigger from CI when a maintainer asks "what MCP servers should we standardize on this quarter?".

Output format?

JSON dataset (one row per MCP server with full GitHub stats + quality score breakdown) plus a Markdown report (mcp-catalog.md). Both downloadable from the Apify Storage tab or via REST API.

Rate limits?

GitHub anonymous API is rate-limited to 60 req/hour. The Actor needs ~1 call per server, so 100 servers ≈ 100 calls. Run during US off-peak or set keywordFilter to narrow scope.

Why not just look at stars?

Stars are gamed and lag adoption. A repo at 1k stars from 2024 with no commits since is worse than a 200-star repo with 50 PRs in the last month. The composite score corrects for that.

What if a list has 1000+ repos?

The Actor stops at maxServers. You can raise it to 500 (the GitHub API rate-limit ceiling for anonymous calls per hour).

Is the data current?

Each run fetches live GitHub data. The dataset is a snapshot at run time — schedule daily/weekly if you need fresh data.

Why use this Actor?

• Pre-install due diligence: before adding an MCP server to your Claude / Cursor / Cline / Aider setup, check if it's maintained
• Curate your own list: drop the JSON into your README to spec-build a higher-signal "awesome" list
• Security teams: find MCP servers without licenses (legal risk) or with abandoned maintainers (supply-chain risk)
• Investors / market analysts: track MCP ecosystem growth over time

How to use

1. (Optional) Filter by keyword (e.g. github, slack, database) to narrow scope
2. Set Max servers (default 100, max 500)
3. (Optional) Add additional Source lists beyond the 3 defaults
4. Click Start
5. Download the dataset or mcp-catalog.md from the Storage tab

Input

• keywordFilter — only include repos whose name contains this keyword (case-insensitive)
• maxServers — stop after N candidates (default 100)
• sourceLists — array of GitHub repos with curated MCP lists (defaults to the 3 largest)

Quality scoring (6 dimensions, max 100 points)

Dimension	Max	What it measures
Stars	30	Log-scaled (avoids over-rewarding mega-repos)
Recency	20	Pushed in last 90d = 20, 180d = 10, 365d = 5
License	15	MIT/Apache/BSD = 15, GPL = 10, none = 0
Description	10	Has substantive repo description
Docs	10	README size proxy (size_kb > 50)
Activity	15	5-200 open issues = healthy; 0 = abandoned

Pricing

Pay-per-event model:

• First run: free trial (up to 50 servers scored)
• Per-server rate: $0.02 per server scored
• Per-catalog (100 servers): $2
• Monthly refresh subscription: $5/month for daily refreshed catalog

Tips

• Set keywordFilter to narrow scope and save API quota — GitHub anonymous API is rate-limited to 60 req/hour
• Run during off-peak (US night) if you hit rate-limits on the GitHub API
• For ongoing tracking, schedule the Actor weekly and diff the dataset to catch new high-quality MCP servers as they appear

Support

Issues / feature requests: open in the Issues tab on the Apify console.

Built by Ian Mu — github.com/ianymu — author of verify-before-stop.