AlaeddineMessadi/gemini-embedding-2-mcp-server
🚀 A powerful Model Context Protocol (MCP) server powered by Gemini Embedding 2. Transforms any local directory into an ultrafast, visually-aware spatial search engine for AI agents! 🔎✨
Platform-specific configuration:
{
"mcpServers": {
"gemini-embedding-2-mcp-server": {
"command": "npx",
"args": [
"-y",
"gemini-embedding-2-mcp-server"
]
}
}
}Add the config above to .claude/settings.json under the mcpServers key.
<div align="center">
<p align="center"> <strong>A powerful Model Context Protocol (MCP) server that transforms any local directory into an ultrafast, visually-aware spatial search engine for AI agents.</strong> </p>
[](https://opensource.org/licenses/MIT) [](https://python.org) [](https://modelcontextprotocol.io/) </div>
---
Connect your local documents, code, images, and videos directly to Claude, Cursor, or VS Code using Google's state-of-the-art gemini-embedding-2-preview model and a strictly local ChromaDB vector database.
| Feature | Description | | :--- | :--- | | 🛡️ Local Privacy | Uses ChromaDB entirely locally (~/.gemini_mcp_db). Your files never go to a 3rd party database. Only raw byte chunks are sent to the Gemini Embedding API. | | 🧠 Enterprise-Grade | Leverages gemini-embedding-2-preview with specialized RETRIEVAL_DOCUMENT Task Types and MRL 768 dimensionality optimization. | | 📸 Ultimate Multimodality | Natively scans, embeds, and retrieves Images (.jpg, .webp), Video (.mp4), and Audio (.mp3, .wav) without extracting text! | | 📄 Visual PDF RAG | Parses PDFs page-by-page as high-quality images. It visually embeds charts, plots, and layout while preserving extracted text for LLM citation. | | 🤖 Agentic Guardrails | Built for autonomous AI agents. Includes an automatic Junk Filter (node_modules, .git), wildcard blacklisting (fnmatch), API exponential backoff, and ghost file pruning. | | ⚡ Smart Deduplication | Pre-calculates MD5 hashes of local files before querying Gemini. Identical, unmodified files bypass the API entirely to save your token
Loading reviews...