AgentRank

Google PageRank for AI agents. A live, daily-updated index of 25,000+ MCP servers and agent tools — scored by real GitHub signals, not just star counts.

👁 npm version
👁 npm downloads
👁 GitHub stars
👁 License: MIT
👁 Last commit

Install in 3 Steps

1. Add to your AI tool:

# Claude Code
claude mcp add agentrank -- npx -y agentrank-mcp-server

2. Ask your AI to find a tool:

"Find me an MCP server for database access"

3. Get ranked, current results — automatically.

That's it. No API key, no config, no prompting. Your AI searches the live index whenever it needs a tool.

Cursor, VS Code, Cline, Claude Desktop, Windsurf →

Why AgentRank?

• Your AI's knowledge is stale. Training data is months old — it can't know if a tool was abandoned last week or if something better shipped yesterday.

• Stars lie. A repo with 2,000 stars and no commits in 18 months isn't production-ready. AgentRank weighs freshness, issue health, and dependents — not just popularity.

• 25,000+ tools, scored daily. The index crawls GitHub nightly. Every recommendation reflects what's happening now.

• Gets smarter with use. Every query surfaces which tools developers actually reach for, sharpening rankings for everyone.

Demo

> search("postgres mcp server")

1. postgres-mcp Score: 91 ★ 2.1k Updated: 2d ago
2. mcp-server-postgres Score: 84 ★ 891 Updated: 5d ago
3. pg-mcp Score: 71 ★ 432 Updated: 12d ago

> lookup("github.com/modelcontextprotocol/servers")

{
 "name": "modelcontextprotocol/servers",
 "agentrank_score": 97,
 "stars": 14200,
 "last_commit": "1 day ago",
 "dependents": 1840,
 "issue_health": 0.91
}

Links

• agentrank-ai.com — Browse the full index

• API Docs — Use the REST API directly

• Blog — MCP server guides and comparisons

• npm: agentrank-mcp-server — The MCP server package

Repo Structure

crawler/ GitHub crawler — finds and indexes repos nightly
scorer/ Scoring engine — 5-signal composite score (0-100)
mcp-server/ MCP server package published to npm
workers/ Cloudflare Workers API
site/ Astro frontend (agentrank-ai.com)

How the Score Works

Five signals, weighted by signal quality:

Scores are recomputed nightly from live GitHub data.

MIT License · Built by @comforteagle

Maintenance

Resources

• NPM Package
• GitHub Repository
Need Help?Related Servers