An offline recommendation engine with 20,000+ AI resources — skills, tools, agents, rules, and MCP servers. Describe what you're building and get ranked, explained results. Works as a standalone CLI or as an MCP server inside Codex, Claude, Cursor, and Windsurf.
npx skillhub-mcp setupNode.js 18+ required. No global install needed.
20,000+ resources bundled locally. Skills, tools, agents, rules, and MCP servers across Cursor, Windsurf, Copilot, Claude, OpenAI, Gemini, and more. No network access — everything is offline.
A TF-IDF search index with multi-signal ranking. Analyzes your task description, matches against technologies, use cases, and categories. Returns scored, explained results — not just keyword matches.
Use it two ways: run commands directly in your terminal (CLI mode), or connect it to your AI client so your assistant can query it automatically (MCP mode).
Run commands directly in your terminal. No setup required — just use npx. Results appear inline with scores, descriptions, and match explanations.
--json for scripting and pipingRegister as an MCP server in your AI client. Your assistant gains tool access and can recommend, search, and inspect resources autonomously.
npx skillhub-mcp setuprecommendDescribe a task in natural language — get ranked tool recommendations with match scores and explanations.
npx skillhub-mcp recommend "build a REST API with auth"searchKeyword search across the full database. Returns matching resources with type, ecosystem, and trust info.
npx skillhub-mcp search "vector database"infoFull details on a specific resource — install command, platforms, trust score, documentation links.
npx skillhub-mcp info "LangChain"statsDatabase overview — total resources, type breakdown, ecosystem distribution.
npx skillhub-mcp statssetupAuto-detect installed AI clients and configure them. Writes the correct MCP config for each one.
npx skillhub-mcp setupdoctorDiagnostic check — Node.js version, npx path, database integrity, client config status.
npx skillhub-mcp doctornpx skillhub-mcp recommend "build a RAG pipeline with LangChain" #1 LangChain ✓ 9.2
Framework for building LLM-powered applications
tool · cross-platform · langchain-ai
→ matches technologies: langchain, python; relevant to: rag, llm
#2 Chroma 8.8
Open-source embedding database for AI
tool · cross-platform · chroma-core
→ matches technologies: embeddings; relevant to: vector-db, rag
#3 LlamaIndex 8.5
Data framework for connecting LLMs to data sources
tool · cross-platform · run-llama
→ matches technologies: python, llm; relevant to: rag, indexingEach result includes a relevance score, match explanation, and resource metadata. Add --json for machine-readable output.
Automatic: run npx skillhub-mcp setup to detect and configure all installed clients. Manual configs below:
# In ~/.codex/config.toml [mcp_servers.skillhub] command = "/opt/homebrew/bin/npx" args = ["-y", "skillhub-mcp@latest"]
Why absolute paths? GUI applications often don't inherit your shell's PATH. The setup wizard resolves the correct absolute path for npx automatically, preventing "command not found" errors.
This package is designed to be auditable and safe to run. Zero network requests, zero runtime dependencies, zero dynamic code execution.
One command configures your AI clients automatically. Or use the CLI directly — no setup needed.
npx skillhub-mcp setup