Package Exports
- skillhub-mcp
- skillhub-mcp/server
- skillhub-mcp/types
Readme
skillhub-mcp
AI resource intelligence for Claude, Cursor, Windsurf, and VS Code. 20,000+ AI skills, tools, agents, rules, and MCP servers — recommended in context.
An MCP server that gives your AI assistant the power to discover, recommend, and explain the right AI tools for whatever you're building.
Install
Claude Desktop
Add to ~/Library/Application Support/Claude/claude_desktop_config.json:
{
"mcpServers": {
"skillhub": {
"command": "npx",
"args": ["-y", "skillhub-mcp"]
}
}
}Restart Claude Desktop. The AI can now recommend AI tools during any conversation.
Cursor
Open Settings → MCP Servers → Add, then:
{
"mcpServers": {
"skillhub": {
"command": "npx",
"args": ["-y", "skillhub-mcp"]
}
}
}Windsurf / VS Code / Any MCP Client
Same pattern — configure your MCP client to run:
npx -y skillhub-mcpGlobal Install (optional)
npm install -g skillhub-mcpNow you can run skillhub-mcp directly, no npx needed.
What Does It Do?
Once configured, your AI assistant gains access to 5 intelligence tools backed by a database of 20,000+ AI resources:
| Tool | What It Does |
|---|---|
recommend |
Analyzes your task and returns ranked recommendations |
search |
Searches resources by name, type, or ecosystem |
get_resource |
Gets full details about a specific resource |
get_setup_guide |
Provides install and configuration instructions |
analyze_stack |
Analyzes your tech stack and suggests complementary AI tools |
You don't call these tools. Your AI calls them automatically when it detects you could benefit from an AI resource.
Examples
"I need to build a RAG pipeline with LangChain"
→ Returns: LangChain, Pinecone, ChromaDB, Weaviate, LlamaIndex RAG components
"Set up Cursor with Next.js best practices"
→ Returns: Next.js Cursor Rules, TypeScript Cursor Rules, React patterns
"I want to run LLMs locally on my machine"
→ Returns: Ollama, llama.cpp, vLLM, LM Studio
"Find MCP servers for database access"
→ Returns: PostgreSQL MCP Server, MongoDB MCP Server, Redis MCP Server
CLI
Also usable as a CLI:
# Recommendations
npx skillhub-mcp recommend "build a RAG pipeline"
# Search
npx skillhub-mcp search "vector database"
# Resource details
npx skillhub-mcp info "Pinecone"
# Database stats
npx skillhub-mcp statsOr if installed globally:
skillhub recommend "deploy ML model with Docker"
skillhub search "MCP server"
skillhub info "LangChain"
skillhub statsDatabase Coverage
| Category | Count |
|---|---|
| Rules (Cursor, Windsurf, Copilot) | 11,588 |
| Instructions | 5,222 |
| Skills / Libraries | 2,012 |
| Tools / SaaS | 967 |
| Agents | 292 |
| MCP Servers | 161 |
| Total | 20,306 |
Ecosystems: Cross-platform, Cursor, GitHub Copilot, Windsurf, OpenAI, Anthropic, Gemini, MCP
How It Works
- Prompt Analysis — Extracts technologies, task categories, and keywords
- TF-IDF Search — Full-text search across 12,000+ indexed terms
- Multi-Signal Ranking — Combines text similarity, tag overlap, type/ecosystem boosting, popularity, and verification quality
- Contextual Results — Ranked, explained, with install guidance
All processing is local. No network requests. No API keys needed.
Verify Installation
After configuring, ask your AI:
"What AI tools would help me build a real-time data pipeline?"
If SkillHub is working, the AI will call the recommend tool and return contextual suggestions.
Or test via CLI:
npx skillhub-mcp statsExpected output: 📊 Database: 20,306 resources
Troubleshooting
Server not showing in Claude Desktop?
- Make sure you restart Claude Desktop after editing config
- Check config file path:
~/Library/Application Support/Claude/claude_desktop_config.json - Verify JSON is valid (no trailing commas)
npx fails?
- Ensure Node.js ≥ 18:
node --version - Try:
npx -y skillhub-mcp
Uninstall
Remove the skillhub entry from your MCP config file and restart the client.
If globally installed: npm uninstall -g skillhub-mcp
License
MIT