RAG MCP Server（RAGDB)

@musoukun

リポジトリ

基本情報

カテゴリ

データベース

トランスポート

stdio

作者

musoukun

RAG MCP Server（RAGDB) について

MCP対応のRAGシステム。Markdownドキュメントをベクトル化し、自然言語で高速検索。LibSQL、Qdrant、PostgreSQLに対応してます。

#ragdb#mcp#search-assistant

RAG MCP Server（RAGDB) の設定

以下の設定を使って、このサーバーを MCP 対応クライアントに追加してください。

{
  "mcpServers": {
    "rag-server": {
      "command": "node",
      "args": [
        "your-ragdb-mcp-path/rag-mcp/dist/index.js"
      ],
      "env": {
        "RAG_DATABASE_TYPE": "libsql",
        "RAG_CONNECTION_URL": "file:your-ragdb-mcp-path/rag-mcp/libsql/rag.db",
        "EMBEDDING_PROVIDER": "google",
        "EMBEDDING_MODEL": "text-embedding-004",
        "EMBEDDING_API_KEY": "your-google-api-key",
        "EMBEDDING_DIMENSIONS": "768",
        "RAG_CHUNK_SIZE": "512",
        "RAG_CHUNK_OVERLAP": "50",
        "RAG_TOP_K": "5",
        "RAG_STRATEGY": "markdown",
        "AUTO_CREATE_INDEXES": "documents,technical",
        "LOG_LEVEL": "info"
      }
    }
  }
}

RAG MCP Server（RAGDB) のツール

ツール一覧はまだありません

この掲載には表示できるツール一覧がまだありません。サーバーにツールがないという意味ではありません。対応機能はプロジェクトのドキュメントをご確認ください。

ツールは README から自動的に抽出されます。メンテナーは ## Tools という見出しの下に記載することで、このタブに反映できます。

概要

What is RAG MCP Server（RAGDB）?

RAG MCP Server（RAGDB） is an MCP server that allows MCP-compatible chat applications to manage and search a RAGDB. It vectorizes Markdown documents, stores them in a vector database, and enables retrieval through natural language questions or semantic similarity.

How to use RAG MCP Server（RAGDB）?

Install by adding a JSON configuration block to your MCP client settings (e.g., VS Code User Settings JSON or Claude Desktop). Set required environment variables for the database type, connection URL, embedding provider, and API key. Then invoke tools such as create_index, add_document, search_documents, or rag_search directly from the chat interface.

Key features of RAG MCP Server（RAGDB）

Save Markdown documents as vector embeddings in RAGDB.
Search documents using natural language questions or semantic similarity.
Choose from three databases: LibSQL (local), Qdrant (cloud), or PostgreSQL.
Batch register multiple Markdown files at once.
Supports OpenAI and Google embedding models.
Configurable chunk size, overlap, and number of search results.

Use cases of RAG MCP Server（RAGDB）

Store internal technical documentation and retrieve it via chat.
Ask questions about project documents during development.
Build a personal knowledge base that can be queried conversationally.
Manage and update document collections with versioning support.

よくある質問

What databases does RAG MCP Server（RAGDB） support?

It supports LibSQL (local file-based, no DB server needed), Qdrant (cloud-ready for large data), and PostgreSQL (tested, for large data). For Qdrant and PostgreSQL, a database server is required.

What embedding providers are supported?

Two providers are supported: OpenAI and Google. You must provide the corresponding API key and set the embedding model name and dimensions via environment variables.

Can I use this server commercially?

Yes, free use, modification, and commercial use are allowed. However, the implementation uses Mastra libraries (licensed under Elastic License 2.0), so directly offering the Mastra Playground as a SaaS service is not permitted.

How do I configure the server for my environment?

Add a JSON block to your MCP client settings with the `mcpServers` entry. Set at least `RAG_DATABASE_TYPE`, `RAG_CONNECTION_URL`, `EMBEDDING_PROVIDER`, and `EMBEDDING_API_KEY`. Optional settings include chunk size, overlap, and auto-create index names.

What tools does RAG MCP Server（RAGDB） provide?

It offers tools such as `create_index`, `delete_index`, `add_document`, `update_document`, `search_documents`, `rag_search`, `advanced_rag_search`, `semantic_similarity_search`, `list_documents`, and `get_rag_info`. Each tool has specific parameters documented in the README.