SkillMesh MCP Server

The intelligent mesh between your codebase and every AI agent.

SkillMesh is a fully local MCP (Model Context Protocol) server that gives AI coding agents the ability to discover, fetch, and install Agent Skills — SKILL.md files that permanently enrich the agent's knowledge of your project.

What Are Agent Skills?

Agent Skills are SKILL.md files that live in your agent's skill directory. Every agent reads all SKILL.md files at the start of every session — skills are always present when you prompt. Installing a skill permanently teaches the agent project-specific rules, patterns, conventions, and best practices.

Features

• 8 MCP Tools for complete skill lifecycle management
• Supports 11 AI Agents: Claude Code, Antigravity, Cursor, Windsurf, Gemini CLI, Codex CLI, Kilo Code, Trae AI, Void Editor, Continue.dev, Cline
• Triple Installation Path: npx command → direct GitHub fetch → LLM generation
• Quality Filtering: Searches skills.sh for skills with ≥50 GitHub stars
• Multi-Provider LLM: Supports Anthropic, Google, and OpenAI for skill generation

Installation

Quick Start

Add SkillMesh to your AI agent's MCP configuration:

{
  "mcpServers": {
    "skillmesh": {
      "command": "npx",
      "args": ["-y", "skillmesh-mcp"],
      "env": {
        "TAVILY_API_KEY": "tvly-your-key-here",
        "ANTHROPIC_API_KEY": "sk-ant-your-key-here",
        "GITHUB_TOKEN": "ghp_your-token-here"
      }
    }
  }
}

Config File Locations

| Agent | Config Path | |-------|-------------| | Claude Code | ~/.claude/claude_mcp_config.json | | Cursor | ~/.cursor/mcp.json | | Windsurf | ~/.codeium/windsurf/mcp_config.json | | Antigravity | Agent-specific config path |

Environment Variables

| Variable | Required | Description | |----------|----------|-------------| | TAVILY_API_KEY | Yes | Web search for finding skills on skills.sh | | ANTHROPIC_API_KEY | No* | LLM for skill generation (preferred) | | GOOGLE_API_KEY | No* | Alternative LLM provider | | OPENAI_API_KEY | No* | Alternative LLM provider | | GITHUB_TOKEN | Recommended | Higher rate limits (60 → 5000 requests/hour) |

*At least one LLM API key required for generate_skill tool.

Available Tools

1. detect_installed_agents

Scans project root and home directory for installed AI coding agents. Returns list of detected agents with their skill directory paths.

2. create_skills_folder

Initializes skill directory structure for selected agents. Creates .claude/skills/, .cursor/skills/, etc.

3. search_skills

Searches skills.sh via Tavily for skills matching a keyword. Fetches GitHub star count and recency. Returns ranked list filtered by quality.

4. fetch_skill_content

Fetches full SKILL.md content from a GitHub repository. Validates YAML frontmatter and required sections.

5. install_skill

Core installation tool with triple-path strategy:

• PATH A: Returns npx skills add command for agent execution
• PATH B: Fetches from GitHub raw URLs, writes directly to disk
• PATH C: Calls LLM to generate skill when none found

6. generate_skill

LLM-powered skill generation. Searches web for best practices, generates a gold-standard SKILL.md file.

7. list_installed_skills

Lists all installed skills across agent directories with metadata.

8. delete_skill

Removes a specific SKILL.md from one or all agent directories.

Usage

Once configured, simply tell your AI agent:

• "Set up skills for this project"
• "Install skills for my codebase"
• "Add relevant skills"
• "Analyze my project and install skills"

The agent will automatically:

1. Detect installed agents
2. Create skill folders
3. Analyze your codebase
4. Search for relevant skills
5. Install or generate skills as needed

Development

# Install dependencies
npm install

# Build
npm run build

# Run tests
npm test

# Development mode
npm run dev

Project Structure

skillmesh-mcp/
├── src/
│   ├── index.ts              # MCP server entry point
│   ├── tools/                # 8 MCP tool implementations
│   │   ├── detect-agents.ts
│   │   ├── create-folder.ts
│   │   ├── search-skills.ts
│   │   ├── fetch-skill.ts
│   │   ├── install-skill.ts
│   │   ├── generate-skill.ts
│   │   ├── list-skills.ts
│   │   └── delete-skill.ts
│   ├── lib/                  # Shared utilities
│   │   ├── agent-paths.ts    # Agent directory mappings
│   │   ├── config.ts         # Config file management
│   │   ├── github.ts         # GitHub API utilities
│   │   ├── llm.ts            # Multi-provider LLM
│   │   ├── skill-parser.ts   # SKILL.md parsing
│   │   ├── skill-sanitizer.ts
│   │   └── tavily.ts         # Web search
│   └── prompts/
│       └── system-prompt.ts
├── tests/
│   └── tools/                # Vitest test files
├── dist/                     # Built output
├── package.json
├── tsconfig.json
└── tsup.config.ts

Supported Agents

| Agent | Project Path | Global Path | |-------|--------------|-------------| | Claude Code | .claude/skills/ | ~/.claude/skills/ | | Antigravity | .agent/skills/ | ~/.gemini/antigravity/global_skills/ | | Cursor | .cursor/skills/ | ~/.cursor/skills/ | | Windsurf | .windsurf/skills/ | ~/.codeium/windsurf/skills/ | | Gemini CLI | .gemini/skills/ | ~/.gemini/skills/ | | Codex CLI | .codex/skills/ | ~/.codex/skills/ | | Kilo Code | .kilocode/skills/ | ~/.kilocode/skills/ | | Trae AI | .trae/skills/ | ~/.trae/skills/ | | Void Editor | .void/skills/ | ~/.void/skills/ | | Continue.dev | .continue/skills/ | ~/.continue/skills/ | | Cline | .cline/skills/ | ~/.cline/skills/ |

License

MIT