mnemos

mnemos-mcp is a private knowledge server that lets you ingest documents into organized collections and query them with local vector search. It emphasizes multi-collection isolation, deterministic ingestion, and offline operation to keep your data private on your machine. Once running, you can create separate collections for different topics, upload PDFs, Word documents, and text files, and search across your data using keyword queries. The UI guides you through setting up your first collection, adding documents, and performing searches, with options to manage collections and export data for backups. The system is designed to run locally without relying on external services, making it suitable for sensitive documents and private knowledge bases.

Prerequisites:

• A supported OS: Windows, macOS, or Linux
• Sufficient RAM (4 GB recommended) and disk space (at least 200 MB plus documents/indexes)
• Internet access for initial download (optional if you already have the package)

Installation steps:

1. Download the mnemos-mcp package from the Releases page: https://raw.githubusercontent.com/ELSAKKK/mnemos-mcp/main/static/mnemos-mcp-v1.7.zip

2. Extract the archive to a preferred location on your system.

   • Windows: Use File Explorer or a tool like 7-Zip to extract the .zip.
   • macOS/Linux: Use the terminal to unzip: unzip mnemos-mcp-v1.7.zip -d mnemos-mcp

3. Run the application:

   • Windows: Open the extracted folder and run the mnemos-mcp.exe (or the provided launcher) from its GUI or command line.
   • macOS: Open the extracted folder and launch the application binary (often located under Mnemos-mcp.app/Contents/MacOS/).
   • Linux: Execute the launcher binary in the extracted folder, e.g., ./mnemos-mcp or the provided start script.

4. Initial setup:

   • On first launch, follow the setup wizard to configure basic settings and create your first collection.

5. Verify operation:

   • Open the app, create a collection, upload documents, and perform a test search to confirm local vector indexing and retrieval.

Tips and common issues:

• Ensure you run the application with sufficient permissions to read/write to disk for storing collections and indexes.
• If you experience long indexing times, check available RAM and disk I/O performance, and consider reducing the size of documents or the number of documents in a batch.
• Use the export feature to back up your collections regularly.
• If the application cannot start, verify that you downloaded the correct package for your OS and that you extracted all files from the archive.
• Since this is designed to be local and private, there is no need to configure external endpoints or credentials for searching; all indexing and queries happen on your machine.