Directory Indexer turns your local directories into an AI-powered knowledge base. It uses an embedding provider (such as Ollama) to create vector representations of your files and stores them in a Qdrant vector database for fast semantic search. Once indexed, your AI assistant can query and retrieve relevant content from your files using natural language, allowing you to ask questions like, Find API authentication examples or Show me error handling patterns, and get results drawn from your actual documents. The MCP server can be launched alongside your AI assistant, and indexing can run independently of the MCP startup, so you don’t have to wait for indexing to finish before using the search capability.

To use it with an AI assistant, configure the MCP with the directory-indexer serve command, index your directories, and then interact with your assistant to perform semantic search over the indexed content. You can organize workspaces for focused searches and customize endpoints and data directories via environment variables in the MCP configuration.

Prerequisites:

• Docker (recommended for Qdrant and Ollama) and/or native installations if desired
• Node.js 18+ installed on your system

1. Install dependencies and set up prerequisites

   • Install Docker from the official site: https://docs.docker.com/get-docker/
   • Install Node.js 18+ from https://nodejs.org/

2. Run Qdrant vector database (Docker option) docker run -d --name qdrant -p 127.0.0.1:6333:6333 -v qdrant_storage:/qdrant/storage qdrant/qdrant

3. Run Ollama embedding service (Docker option or native installation) docker run -d --name ollama -p 127.0.0.1:11434:11434 -v ollama:/root/.ollama ollama/ollama

   Pull embedding model

   docker exec ollama ollama pull nomic-embed-text

4. Install the Directory Indexer package (via npm) npm install -g directory-indexer@latest

5. Index your directories (example) npx directory-indexer@latest index ./WorkNotes ./Projects

6. Start the MCP server to serve the index (example) npx directory-indexer@latest serve

7. Optional: configure MCP integration with your AI assistant using the provided MCP config example: { "mcpServers": { "directory-indexer": { "command": "npx", "args": ["directory-indexer@latest", "serve"] } } }

Environment variables and configuration tips:

• DIRECTORY_INDEXER_DATA_DIR: path where the index data will be stored (default is ~/.directory-indexer)
• QDRANT_ENDPOINT: endpoint URL for the Qdrant vector database (default http://localhost:6333)
• OLLAMA_ENDPOINT: endpoint URL for the Ollama embedding service (default http://localhost:11434)

Common issues:

• If indexing runs slowly, ensure Ollama and Qdrant are reachable at their endpoints and that the embedding model is available.
• On Windows, you may need to adapt file paths for workspaces and ensure MCP config uses proper escaping.
• If the MCP server fails to start, check that the port mappings (6333 for Qdrant and 11434 for Ollama) are not in use by other processes.

Workspaces and advanced usage:

• You can define multiple workspaces to constrain searches to specific directories.
• Use the provided examples to customize commands and environment variables for your setup.