Local FAISS

v0.2.0-rc.3Knowledge & Memorystable

Provides local vector database functionality using FAISS for document ingestion, semantic search, and Retrieval-Augmented Generation (RAG) applications with persistent storage and customizable embedding models.

agentic-aiai-agentsai-toolsfaissllm-tools
Share:
30
Stars
0
Downloads
0
Weekly
0/5

What is Local FAISS?

Local FAISS is a Model Context Protocol (MCP) server that allows AI assistants like Claude, Cursor, and VS Code to provides local vector database functionality using faiss for document ingestion, semantic search, and retrieval-augmented generation (rag) applications with persistent storage and customizable embeddi...

Provides local vector database functionality using FAISS for document ingestion, semantic search, and Retrieval-Augmented Generation (RAG) applications with persistent storage and customizable embedding models.

This server falls under the Knowledge & Memory category on MCPgee, the world's largest MCP server directory with 33,000+ servers.

Features

  • Provides local vector database functionality using FAISS for

Use Cases

Build local vector databases for semantic search without external APIs.
Implement Retrieval-Augmented Generation for document-based AI applications.
Store and retrieve embeddings with customizable embedding models.
nonatofabio

Maintainer

LicenseMIT License
Languagepython
Versionv0.2.0-rc.3
UpdatedMay 8, 2026
Statushealthy
Maintenanceactive

Works with

ClaudeOpenAIwindowsmacoslinux

Installation

PIP

pip install local-faiss-mcp

Manual Installation

pip install local-faiss-mcp

Configuration

Configuration Details

Config File

claude_desktop_config.json

Performance

Response Metrics

Response Time< 200ms
ThroughputMedium

Resource Usage

Memory UsageLow
CPU UsageLow

How to Set Up and Use Local FAISS

Local FAISS MCP Server provides a fully local vector database for AI assistants using Facebook's FAISS library, enabling document ingestion, semantic search, and Retrieval-Augmented Generation (RAG) without sending data to external APIs. It supports ingesting PDFs, text files, Markdown, and more, and persists the vector index to disk so your knowledge base survives between sessions. Developers use it to build private RAG applications where documents contain sensitive information that must not leave the local machine.

Prerequisites

  • Python 3.9+ installed
  • pip for installing the local-faiss-mcp package
  • Optional: pandoc installed for DOCX, HTML, and EPUB file support
  • An MCP-compatible client such as Claude Desktop or Claude Code
1

Install from PyPI

Install the local-faiss-mcp package with pip. This installs FAISS, sentence-transformers, and all required dependencies.

pip install local-faiss-mcp
2

Choose an index storage directory

Decide where you want the FAISS index files to be stored on disk. This directory will persist your vector database between sessions. Create it if it does not exist.

mkdir -p ~/faiss-index
3

Run the server to verify it starts

Test the server with your chosen index directory and the default embedding model (all-MiniLM-L6-v2) to confirm everything is installed correctly.

python -m local_faiss_mcp --index-dir ~/faiss-index
4

Add the server to your MCP client config

Register local-faiss-mcp in your Claude Desktop or Claude Code configuration, specifying the index directory and optionally a custom embedding model.

{
  "mcpServers": {
    "local-faiss": {
      "command": "python",
      "args": [
        "-m", "local_faiss_mcp",
        "--index-dir", "/Users/you/faiss-index",
        "--embed", "all-MiniLM-L6-v2"
      ]
    }
  }
}
5

Ingest your first document

Use the ingest_document tool to add a document to the vector store. You can provide a file path or raw text content. The document is chunked, embedded, and stored in FAISS.

6

Query the vector store

Use the query_rag_store tool to perform semantic search over your ingested documents. Results are returned as the most relevant chunks with optional re-ranking.

Local FAISS Examples

Client configuration

Claude Desktop configuration for Local FAISS MCP server with a custom index directory and re-ranking model.

{
  "mcpServers": {
    "local-faiss": {
      "command": "python",
      "args": [
        "-m", "local_faiss_mcp",
        "--index-dir", "/Users/you/faiss-index",
        "--embed", "all-MiniLM-L6-v2",
        "--rerank", "cross-encoder/ms-marco-MiniLM-L-6-v2"
      ]
    }
  }
}

Prompts to try

Example prompts for ingesting documents and querying the local vector store.

- "Ingest the file /Users/me/docs/architecture.pdf into the vector store"
- "Search the vector store for information about authentication flows"
- "What does my documentation say about error handling?"
- "Summarize the key topics covered across all ingested documents"
- "Find the most relevant chunks about database migrations in my docs"

Troubleshooting Local FAISS

ImportError for faiss or sentence_transformers on install

On Apple Silicon Macs, FAISS may need to be installed separately: 'pip install faiss-cpu'. On Linux, use 'pip install faiss-cpu' for CPU-only or 'faiss-gpu' for GPU support. Then retry 'pip install local-faiss-mcp'.

DOCX or HTML files fail to ingest

These formats require pandoc as an external dependency. Install pandoc from https://pandoc.org/installing.html and ensure the 'pandoc' command is available in your PATH. PDF, TXT, and MD files work without pandoc.

Semantic search returns irrelevant results

The default embedding model (all-MiniLM-L6-v2) is fast but general-purpose. For domain-specific corpora, try a more specialized embedding model via the --embed argument. Adding the --rerank flag with a cross-encoder model significantly improves result relevance.

Frequently Asked Questions about Local FAISS

What is Local FAISS?

Local FAISS is a Model Context Protocol (MCP) server that provides local vector database functionality using faiss for document ingestion, semantic search, and retrieval-augmented generation (rag) applications with persistent storage and customizable embedding models. It connects AI assistants to external tools and data sources through a standardized interface.

How do I install Local FAISS?

Install via pip with: pip install local-faiss-mcp. Then configure your AI client to connect to this MCP server.

Which AI clients work with Local FAISS?

Local FAISS works with all major MCP-compatible AI clients including Claude Desktop, Claude Code, Cursor, VS Code (GitHub Copilot), Windsurf, and Cline.

Is Local FAISS free to use?

Yes, Local FAISS is open source and available under the MIT License license. You can use it freely in both personal and commercial projects.

Browse More Knowledge & Memory MCP Servers

Explore all knowledge & memory servers available in the MCPgee directory. Each server includes setup guides for Claude, Cursor, and VS Code.

Quick Config Preview

{ "mcpServers": { "local-faiss-mcp-server": { "command": "pip", "args": ["install", "local-faiss-mcp"] } } }

Add this to your claude_desktop_config.json or .cursor/mcp.json

Read the full setup guide →

Ready to use Local FAISS?

Browse our complete directory of 33,000+ MCP servers, read setup guides for your editor, and start building with the Model Context Protocol.

33,000+ ServersFree & Open SourceStep-by-Step Guides