ChromaDB

A Model Context Protocol (MCP) server that gives AI assistants persistent memory through ChromaDB vector storage.Now with EXIF extraction, Watch Folders, and Duplicate Detection- the ultimate tool for creators!

- Persistent AI Memory: Your AI assistant remembers past conversations and solutions
- Vector Search: Find similar code patterns, configurations, and documentation instantly
- Local First: Run everything on your own hardware, no cloud dependencies

- Fast Batch Ingest: Process entire directories in seconds (500+ files)
- 77 File Types: Photos, CAD, documents, data files, code
- Quick Load/Unload: Temporary collections for rapid workflows
- Export/Import: Backup and transfer collections as JSON

- EXIF Extraction: Camera, lens, exposure, GPS location, date taken
- Search by Camera: "Find photos shot with my Canon 5D"
- Search by Location: GPS coordinates embedded and searchable
- Search by Date: "Find photos from vacation 2024"

- Auto-Ingest: Drop files in watched folders, auto-add to ChromaDB
- Hands-Free: Perfect for incoming photo dumps, downloads
- Filter by Type: Watch only for specific file types

- Find Duplicates: Hash-based detection across directories
- Reclaim Space: See exactly how much space duplicates waste
- Compare Files: Check if two files are identical
- Perceptual Hashing: Find similar (not just identical) images

- Bun(JavaScript runtime)
- Docker(for ChromaDB)
- Claude Desktop(or any MCP client)

git clone https://github.com//vespo92/chromadblocal-mcp-server.git cd chromadb-mcp-server

docker run -d \ --name chromadb-local \ -p 8001:8000 \ -v ~/chromadb-data:/chroma/chroma \ -e IS_PERSISTENT=TRUE \ chromadb/chroma:latest

{ "mcpServers": { "chromadb-context": { "command": "bun", "args": ["run", "/path/to/chromadb-mcp-server/index.js"], "env": { "CHROMADB_URL": "http://localhost:8001" } } } }

Restart Claude Desktopand start building your knowledge base!

Once configured, interact naturally with your AI:

- "Store this Docker configuration in ChromaDB for future reference"
- "Save this React component pattern with tags: hooks, authentication"
- "Remember this solution for GPU passthrough issues"

- "Search ChromaDB for Python async examples"
- "Find similar component patterns to this one"
- "What solutions do we have for Docker networking issues?"

- "Add this API documentation to the project_docs collection"
- "Store these test patterns for our testing suite"

The killer feature! Process massive amounts of files instantly for AI-powered search and retrieval.

Perfect for "load, process, discard" workflows:

You: "Quick load my photos from /home/photos/vacation2024" AI: Creates temp collection, ingests 500 photos in seconds You: "Find photos with mountains or beaches" AI: Returns matching photos with metadata You: "Unload the collection" AI: Cleans up, frees memory

"Scan /projects/cad-files to see what's there" "Batch ingest all STL files from /3d-prints into the 'print_library' collection" "Quick load my Downloads folder, find anything mentioning 'invoice'" "Export the photo_archive collection to backup.json" "Import backup.json into a new collection called 'restored_photos'"

- Quick Load: ~200 files in 2-3 seconds
- Batch Ingest: ~500 files in 5-10 seconds (with full metadata)
- Concurrent Processing: 10-20 parallel file operations
- No external dependencies: Pure JavaScript/Bun processing

Search for relevant information across collections

Parameters: - query: Search query - collection: (optional) Specific collection to search - limit: (optional) Number of results

Parameters: - content: The content to store - metadata: Tags, categories, descriptions - collection: Target collection

List all available collections and their metadata

Find code patterns similar to provided example

Preview files in a directory before ingesting

Parameters: - path: Directory to scan - categories: Filter by type (images, cad, documents, data, code) - extensions: Filter by extension (.jpg, .stl, etc.) - recursive: Include subdirectories (default: true)

Bulk ingest files into ChromaDB with full metadata

Parameters: - path: Source directory - collection: Target collection name - categories: File types to include - max_files: Limit number of files

🚀 FAST: Rapidly load files for temporary processing

Parameters: - path: Directory to load - name: Collection name (auto-generated if omitted) - categories: File types to include

Delete a collection (cleanup after quick_load)

Parameters: - collection: Name of collection to delete

Parameters: - collection: Collection to export - output_path: File path for JSON output

Parameters: - input_path: JSON file to import - collection: Override collection name - overwrite: Delete existing first (default: false)

Parameters: - collection: Collection name

Ingest a single file with metadata extraction

Parameters: - path: File to ingest - collection: Target collection

Extract detailed EXIF metadata from photos

Parameters: - path: Path to JPEG or TIFF image Returns: Camera, lens, exposure, GPS, date taken

Start auto-ingesting new files from a folder

Parameters: - path: Folder to watch - collection: Target collection (default: auto_ingest) - categories: File types to watch - include_exif: Extract EXIF from photos (default: true)

Parameters: - path: Folder to stop watching

Parameters: - path: Directory to scan - hash_method: "partial" (fast), "full" (thorough), "perceptual" (images) - categories: File types to check Returns: Duplicate groups with wasted space info

Parameters: - file1: First file path - file2: Second file path

Find duplicate entries in a ChromaDB collection

Parameters: - collection: Collection name

CHROMADB_URL=http://localhost:8001 # ChromaDB server URL

Add new collections insetup-home-collections.js:

await createCollection('ml_experiments', { description: 'Machine learning experiments and results' });

chromadb-mcp-server/ ├── index.js # MCP server with 22 tools ├── batch-processor.js # Fast batch file processing engine ├── exif-extractor.js # EXIF metadata extraction for photos ├── watch-folder.js # Auto-ingest watch folder system ├── duplicate-detector.js # Duplicate file detection ├── setup-home-collections.js # Collection initialization ├── test-chromadb.js # Connection test script ├── test-mcp.js # MCP functionality test ├── test-batch-processor.js # Batch processing tests ├── HOME-AI-SETUP.md # Detailed setup guide ├── package.json # Project dependencies └── README.md # This file

Contributions are welcome! Please feel free to submit a Pull Request. For major changes, please open an issue first to discuss what you would like to change.

This project is licensed under the MIT License - see theLICENSEfile for details.

- Anthropicfor the MCP specification
- Chromafor the excellent vector database
- The open-source community for inspiration and support

A production-ready Model Context Protocol (MCP) server that provides a powerful, vector-native memory bank for AI agents. Built with the Protocol-Lattice Go Agent Framework, this server offers persistent, searchable, and shareable memory with multiple database backends.