Snippets MCP

• MCP server for storing, searching, and managing code snippets using semantic search and traditional keyword matching.
• Just tell your coding agent (claude code, cursor, cline, opencode, etc.) to save a certain snippet. That's it.
• When needed, just tell it to search for code snippets related to: your query.

Features

• Semantic search using AI embeddings to find snippets by meaning, not just keywords
• Hybrid search combining semantic similarity and keyword matching
• Automatic programming language detection
• Tag-based organization and filtering
• Date range filtering
• No database needed. JSON based storage.
• Vector embeddings cached for fast retrieval

Installation

npm install @freakynit/snippets-mcp

Available Tools

add-snippet

Adds a new code snippet to the database.

Parameters:

• code (string, required) - The code content
• tags (array, optional) - Array of tag strings
• language (string, optional) - Programming language (auto-detected if not provided)
• description (string, optional) - Text description for better semantic search

search-snippets

Searches snippets using hybrid semantic and keyword matching.

Parameters:

• query (string, optional) - Natural language search query
• tags (array, optional) - Filter by specific tags (AND logic)
• language (string, optional) - Filter by programming language
• dateStart (ISO date string, optional) - Filter by creation date start
• dateEnd (ISO date string, optional) - Filter by creation date end
• limit (number, optional) - Maximum results to return (default: 10)

update-snippet

Updates an existing snippet. Re-generates embeddings if code, tags, or description change.

Parameters:

• id (string, required) - Snippet ID
• updates (object) - Object containing fields to update (code, tags, language, description)

delete-snippet

Deletes a snippet from the database.

Parameters:

• id (string, required) - Snippet ID

get-snippet

Retrieves a single snippet by ID.

Parameters:

• id (string, required) - Snippet ID

Environment Variables

1. SNIPPETS_FILE_PATH: Optional, Full path to file to save snippets and embeddings in. Defaults to ~/.snippets-mcp-db.json.

How It Works

The library uses a hybrid search approach:

1. Semantic Search (70% weight) - Uses the all-MiniLM-L6-v2 model to perform vector searh against embeddings generated off code, description, tags and language.
2. Keyword Matching (30% weight) - Traditional text matching for exact term matches based on code and tags.
3. Hard Filters - Applied first to narrow results by tags, language, and date range.

Embeddings are generated once when adding/updating snippets and cached for fast retrieval.

Storage

Snippets are stored in a JSON file specified by environment variable SNIPPETS_FILE_PATH, or at default path: ~/.snippets-mcp-db.json with the following structure:

{
  "id": "uuid",
  "code": "string",
  "language": "string",
  "tags": ["array"],
  "description": "string",
  "embedding": [/* vector array */],
  "createdAt": "ISO date",
  "updatedAt": "ISO date"
}

Configuring using mcpServers json

For Mac and Linux

{
  "mcpServers": {
    "snippets-mcp": {
      "command": "npx",
      "args": ["-y", "@freakynit/snippets-mcp@latest"],
      "env": {
        "SNIPPETS_FILE_PATH": "Optional... path to save snippets and embeddings in.. should have .json extension"
      }
    }
  }
}

For Windows

{
  "mcpServers": {
    "snippets-mcp": {
      "command": "cmd",
      "args": ["/k", "npx", "-y", "@freakynit/snippets-mcp@latest"],
      "env": {
        "SNIPPETS_FILE_PATH": "Optional... path to save snippets and embeddings in.. should have .json extension"
      }
    }
  }
}

License

MIT