Agent Skills & MCP Server Directory: skills.sh & Smithery

G2 for AI agent skills. Find the best agent skills (Claude Code, Codex, Cursor, GitHub Copilot, Gemini CLI, Windsurf) and MCP servers in one run. Each one is ranked by real adoption data: installs, 8-week install trend, Smithery usage, npm downloads and GitHub stars. Security audit verdicts from Gen Agent Trust Hub, Socket and Snyk act as the "reviews".

Sources:

• skills.sh, the open agent-skills directory: all-time, trending and hot leaderboards (top 600 each) and each skill's page.
• Smithery, the largest MCP server marketplace: usage count, verified flag, remote/hosted status and the full tool list of each server.
• The official MCP Registry (registry.modelcontextprotocol.io): newly published servers, packages (npm/PyPI/Docker), remotes, repositories, plus npm weekly downloads.

What you can do with it

• Pick the right skill or MCP server: compare options for a task (postgres, browser, stripe, seo) by adoption and security, before you install anything.
• Track the agent ecosystem: the top and fastest-growing skills each week, and new MCP servers as they are published.
• Market and competitor research: see who publishes the most-installed skills and servers (Vercel, Anthropic, Supabase, Microsoft...) and how fast they grow.
• Security review: filter skills whose audits returned warn or fail before rolling them out to a team.
• Build a directory, newsletter or dashboard: clean JSON with stable IDs, ready for Sheets, Airtable or a database.

How to use

1. Optionally add search terms, e.g. postgres, react, browser. Leave empty for the top-ranked lists.
2. Choose sources and the skills.sh leaderboard (all-time, trending or hot).
3. Click Start, then export JSON, CSV or Excel, or connect Make, Zapier, n8n or Slack.

With an empty input you get the top 50 all-time skills (with details), the top 50 Smithery servers by usage (with tools), and the 50 newest MCP Registry servers from the last 14 days.

Input example

{
  "searchTerms": ["postgres", "supabase"],
  "sources": ["skills-sh", "smithery", "mcp-registry"],
  "skillsView": "all-time",
  "includeSkillDetails": true,
  "includeMcpTools": true,
  "maxItemsPerSource": 25
}

Output

Agent skill (skills.sh):

{
  "itemType": "agentSkill", "source": "skills.sh", "leaderboard": "all-time", "rank": 1,
  "id": "vercel-labs/skills/find-skills", "name": "find-skills", "owner": "vercel-labs",
  "repositoryUrl": "https://github.com/vercel-labs/skills",
  "url": "https://www.skills.sh/vercel-labs/skills/find-skills",
  "installs": 4017447, "isOfficial": true,
  "weeklyInstalls": [102581, 101561, 101317, 105144, 103601, 107199, 110524, 121189],
  "weeklyInstallsLatest": 121189, "weeklyInstallsChangePct": 9.6,
  "description": "Helps users discover and install agent skills when they ask questions like...",
  "summary": "Discover and install specialized agent skills from the open ecosystem...",
  "tags": ["Agent workflows"], "githubStars": 36200, "firstSeen": "2026-01-26",
  "securityAudits": [
    { "provider": "Gen Agent Trust Hub", "status": "pass" },
    { "provider": "Socket", "status": "pass" },
    { "provider": "Snyk", "status": "warn" }
  ],
  "securityVerdict": "warn",
  "installCommand": "npx skills add https://github.com/vercel-labs/skills --skill find-skills"
}

MCP server (Smithery):

{
  "itemType": "mcpServer", "source": "smithery", "rank": 1,
  "id": "brave", "name": "Brave Search", "description": "Search the web with Brave's independent index...",
  "url": "https://smithery.ai/server/brave", "useCount": 87579, "verified": true, "remote": true,
  "toolCount": 8, "tools": [{ "name": "brave_web_search", "description": "Performs web searches..." }],
  "connectionTypes": ["http"]
}

MCP server (official MCP Registry):

{
  "itemType": "mcpServer", "source": "mcp-registry", "rank": 1,
  "id": "com.supabase/mcp", "name": "Supabase", "description": "MCP server for interacting with the Supabase platform",
  "version": "0.14.0", "url": "https://supabase.com/mcp", "repositoryUrl": "https://github.com/supabase/mcp",
  "remotes": [{ "type": "streamable-http", "url": "https://mcp.supabase.com/mcp" }],
  "packages": [{ "registryType": "npm", "identifier": "@supabase/mcp-server-supabase", "version": "0.14.0", "transport": "stdio" }],
  "npmWeeklyDownloads": 68012, "status": "active", "publishedAt": "2026-10-14T15:34:03.096613Z"
}

securityVerdict combines the audits: pass (all pass), warn (at least one warning), fail (at least one fail or high risk) or unaudited.

How much does it cost?

Pay per event: $0.005 per run + $0.0015 per result ($1.50 per 1,000), with details included. No compute charges.

You get	You pay (Starter/Bronze plan)
30 results (10 per source)	$0.05
150 results (default, 50 per source)	$0.23
1,000 results	$1.51

Apify's free plan includes $5 of usage every month, and free-plan users pay half price. Paid plans get volume discounts (Silver -10%, Gold and above -20%).

Use cases

• Developers and AI engineers: choose a skill or MCP server for Claude Code, Cursor or Codex with numbers, not hype.
• Platform and security teams: keep an approved list and flag skills with failing audits.
• Dev-tool companies and DevRel: track your skill's and MCP server's installs against competitors, every week.
• Newsletter writers and directory builders: "top new MCP servers this week" and "fastest-growing agent skills" in one scheduled run.
• AI agents: call it through the Apify MCP server to answer "which MCP server should I use for X?".

Run it from code (API, Python, JavaScript)

curl -X POST "https://api.apify.com/v2/acts/api-co~agent-skills-mcp-directory/run-sync-get-dataset-items?token=YOUR_APIFY_TOKEN" \
  -H "Content-Type: application/json" -d '{"searchTerms": ["browser"], "maxItemsPerSource": 10}'

from apify_client import ApifyClient
client = ApifyClient("YOUR_APIFY_TOKEN")
run = client.actor("api-co/agent-skills-mcp-directory").call(run_input={"sources": ["skills-sh"], "skillsView": "trending"})
for s in client.dataset(run["defaultDatasetId"]).iterate_items():
    print(s["rank"], s["id"], s.get("installsRecent"), s.get("securityVerdict"))

import { ApifyClient } from 'apify-client';
const client = new ApifyClient({ token: 'YOUR_APIFY_TOKEN' });
const run = await client.actor('api-co/agent-skills-mcp-directory').call({ searchTerms: ['postgres'] });
const { items } = await client.dataset(run.defaultDatasetId).listItems();

FAQ

What is an agent skill? A folder with a SKILL.md file (plus optional scripts) that teaches an AI coding agent how to do a task. Claude Code, Codex, Cursor, Copilot, Gemini CLI and others can load them. skills.sh tracks installs made with npx skills add.

How does search work? On skills.sh, terms are matched against the name, repository and (with details on) the description, summary and topics of skills on the selected leaderboard (top 600). Smithery and the MCP Registry are searched with their own search APIs.

What do the leaderboards mean? All-time ranks by total installs. Trending ranks by recent growth and reports installsRecent. Hot compares the last hour with the same hour yesterday (installsLastHour, installsSameHourYesterday, hourlyChange).

Why don't MCP Registry servers have a usage number? The official registry does not publish usage. For npm packages the Actor adds npmWeeklyDownloads, and with a GitHub token it adds stars, last push date, license and archived status.

Are the security audits mine to rely on? They come from skills.sh's audit partners and are a useful first filter. Always read a skill's SKILL.md and source before installing it on a machine that matters.

Is this allowed? The Actor reads public pages that robots.txt allows (it does not use skills.sh's /api/ paths) and the public Smithery and MCP Registry APIs, with polite rate limits. Skill content belongs to its authors under their repositories' licenses.

Related Actors

• AI Bot Analytics & AI Crawler Checker: see which AI agents visit your site.
• AI Model Leaderboard: LMArena rankings, prices and new model releases.
• Hacker News Scraper: what developers are discussing.