A Model Context Protocol (MCP) server that provides intelligent search capabilities for discovering relevant Claude Agent Skills using vector embeddings and semantic similarity. This server implements the same progressive disclosure architecture that Anthropic describes in their Agent Skills engineering blog, making specialized skills available to any MCP-compatible AI application. An open-source
Add this skill
npx mdskills install K-Dense-AI/claude-skills-mcpWell-architected MCP server with semantic search, progressive disclosure, and two-package design for optimal performance
1# Claude Skills MCP Server23[](https://github.com/K-Dense-AI/claude-skills-mcp/actions/workflows/test.yml)4[](https://www.python.org/downloads/)5[](https://opensource.org/licenses/Apache-2.0)6[](https://github.com/astral-sh/ruff)7[](https://badge.fury.io/py/claude-skills-mcp)89> **Use [Claude's powerful new Skills system](https://www.anthropic.com/news/skills) with ANY AI model or coding assistant** - including Cursor, Codex, GPT-5, Gemini, and more. This MCP server brings Anthropic's Agent Skills framework to the entire AI ecosystem through the Model Context Protocol.1011A Model Context Protocol (MCP) server that provides intelligent search capabilities for discovering relevant Claude Agent Skills using vector embeddings and semantic similarity. This server implements the same progressive disclosure architecture that Anthropic describes in their [Agent Skills engineering blog](https://www.anthropic.com/engineering/equipping-agents-for-the-real-world-with-agent-skills), making specialized skills available to any MCP-compatible AI application.1213**An open-source project by [K-Dense AI](https://k-dense.ai)** - creators of autonomous AI scientists for scientific research.1415This MCP server enables any MCP-compatible AI assistant to intelligently search and retrieve skills from our curated [Claude Scientific Skills](https://github.com/K-Dense-AI/claude-scientific-skills) repository and other skill sources like the [Official Claude Skills](https://github.com/anthropics/skills).1617<a href="https://cursor.com/en-US/install-mcp?name=claude-skills-mcp&config=eyJjb21tYW5kIjoidXZ4IGNsYXVkZS1za2lsbHMtbWNwIn0%3D">18 <picture>19 <source srcset="https://cursor.com/deeplink/mcp-install-light.svg" media="(prefers-color-scheme: dark)">20 <source srcset="https://cursor.com/deeplink/mcp-install-dark.svg" media="(prefers-color-scheme: light)">21 <img src="https://cursor.com/deeplink/mcp-install-dark.svg" alt="Install MCP Server" style="height:2.7em;"/>22 </picture>23</a>2425## Demo26272829*Semantic search and progressive loading of Claude Agent Skills in Cursor*3031## Highlights3233- **Two-Package Architecture**: Lightweight frontend (~15 MB) starts instantly; backend (~250 MB) downloads in background34- **No Cursor Timeout**: Frontend responds in <5 seconds, solving the timeout issue35- **Semantic Search**: Vector embeddings for intelligent skill discovery36- **Progressive Disclosure**: Multi-level skill loading (metadata → full content → files)37- **Zero Configuration**: Works out of the box with curated skills38- **Multi-Source**: Load from GitHub repositories and local directories39- **Fast & Local**: No API keys needed, with automatic GitHub caching40- **Configurable**: Customize sources, models, and content limits4142## Quick Start4344### For Cursor Users4546Add through the [Cursor Directory](https://cursor.directory/mcp/claude-skills-mcp), or add to your Cursor config (`~/.cursor/mcp.json`):4748```json49{50 "mcpServers": {51 "claude-skills": {52 "command": "uvx",53 "args": ["claude-skills-mcp"]54 }55 }56}57```5859The frontend starts instantly and displays tools, automatically downloading and starting the backend in the background (~60-120s due to RAG dependencies, one-time). Subsequent uses are instant.6061### Using uvx (Standalone)6263Run the server with default configuration:6465```bash66uvx claude-skills-mcp67```6869This starts the lightweight frontend which auto-downloads the backend and loads ~90 skills from Anthropic's official skills repository and K-Dense AI's scientific skills collection.7071### With Custom Configuration7273```bash74# 1. Print the default configuration75uvx claude-skills-mcp --example-config > config.json7677# 2. Edit config.json to your needs7879# 3. Run with your custom configuration80uvx claude-skills-mcp --config config.json81```8283## Documentation8485- **[Getting Started](docs/getting-started.md)** - Installation, Cursor setup, CLI usage, and troubleshooting86- **[Architecture Guide](docs/architecture.md)** - Two-package design, data flow, and components87- **[API Documentation](docs/api.md)** - Tool parameters, examples, and best practices88- **[Usage Examples](docs/usage.md)** - Advanced configuration, real-world use cases, and custom skill creation89- **[Testing Guide](docs/testing.md)** - Complete testing instructions, CI/CD, and coverage analysis9091## MCP Tools9293The server provides three tools for working with Claude Agent Skills:94951. **`find_helpful_skills`** - Semantic search for relevant skills based on task description962. **`read_skill_document`** - Retrieve specific files (scripts, data, references) from skills973. **`list_skills`** - View complete inventory of all loaded skills (for exploration/debugging)9899See [API Documentation](docs/api.md) for detailed parameters, examples, and best practices.100101## Architecture (v1.0.0)102103The system uses a **two-package architecture** for optimal performance:104105- **Frontend** ([`claude-skills-mcp`](https://pypi.org/project/claude-skills-mcp/)): Lightweight proxy (~15 MB)106 - Starts instantly (<5 seconds) ✅ **No Cursor timeout!**107 - Auto-downloads backend on first use108 - MCP server (stdio) for Cursor109110- **Backend** ([`claude-skills-mcp-backend`](https://pypi.org/project/claude-skills-mcp-backend/)): Heavy server (~250 MB)111 - Vector search with PyTorch & sentence-transformers112 - MCP server (streamable HTTP)113 - Auto-installed by frontend OR deployable standalone114115**Benefits:**116- ✅ Solves Cursor timeout issue (frontend starts instantly)117- ✅ Same simple user experience (`uvx claude-skills-mcp`)118- ✅ Backend downloads in background (doesn't block Cursor)119- ✅ Can connect to remote hosted backend (no local install needed)120121See [Architecture Guide](docs/architecture.md) for detailed design and data flow.122123## Skill Sources124125Load skills from **GitHub repositories** (direct skills or Claude Code plugins) or **local directories**.126127By default, loads from:128- [Official Anthropic Skills](https://github.com/anthropics/skills) - 15 diverse skills for documents, presentations, web artifacts, and more129- [K-Dense AI Scientific Skills](https://github.com/K-Dense-AI/claude-scientific-skills) - 78+ specialized skills for bioinformatics, cheminformatics, and scientific analysis130- Local directory `~/.claude/skills` (if it exists)131132## Contributing133134Contributions are welcome! To contribute:1351361. **Report issues**: [Open an issue](https://github.com/K-Dense-AI/claude-skills-mcp/issues) for bugs or feature requests1372. **Submit PRs**: Fork, create a feature branch, ensure tests pass (`uv run pytest tests/`), then submit1383. **Code style**: Run `uvx ruff check src/` before committing1394. **Add tests**: New features should include tests140141### Development142143**Version Management**: This monorepo uses a centralized version system:144- Edit the `VERSION` file at the repo root to bump the version145- Run `python3 scripts/sync-version.py` to sync all references (or use `--check` to verify)146- The `scripts/build-all.sh` script automatically syncs versions before building147148For questions, email [orion.li@k-dense.ai](mailto:orion.li@k-dense.ai)149150## Join Our Community! 🚀151152**We'd love to have you in our Slack community!** Connect with other users, share tips and tricks, get help with your skills, and be the first to know about new features and updates.153154👉 **[Join the K-Dense Community on Slack](https://join.slack.com/t/k-densecommunity/shared_invite/zt-3iajtyls1-EwmkwIZk0g_o74311Tkf5g)** 👈155156Whether you're building custom skills, integrating with different AI models, or just exploring the possibilities of Agent Skills, our community is here to support you!157158## Learn More159160- [Agent Skills Documentation](https://docs.claude.com/en/docs/claude-code/skills) - Official Anthropic documentation on the Skills format161- [Agent Skills Blog Post](https://www.anthropic.com/news/skills) - Announcement and overview162- [Model Context Protocol](https://modelcontextprotocol.io/) - The protocol that makes cross-platform Skills possible163- [Engineering Blog: Equipping Agents for the Real World](https://www.anthropic.com/engineering/equipping-agents-for-the-real-world-with-agent-skills) - Technical deep-dive on the Skills architecture164165## License166167This project is licensed under the [Apache License 2.0](LICENSE).168169Copyright 2025 K-Dense AI (https://k-dense.ai)170171## Star History172173[](https://www.star-history.com/#K-Dense-AI/claude-skills-mcp&type=date&legend=top-left)
Full transparency — inspect the skill content before installing.