ChromaDB

This MCP server provides persistent memory for AI assistants by storing context and metadata in a local ChromaDB vector store. It supports fast batch ingestion, EXIF data from photos, watch folders for auto-ingest, and duplicate detection to optimize storage. You can build rich, searchable knowledge bases that your AI can reference across conversations and sessions.

How to use

You interact with the MCP server through an MCP client to store, search, and retrieve contextual information. Use it to capture code patterns, configurations, and documentation, then pull back relevant results when building solutions or debugging issues. Leverage batch processing to ingest large collections, watch folders to automatically add new files, and duplicate detection to reclaim space. Use the search and store tools to organize knowledge with tags, and then retrieve the most relevant items by comparing patterns and similarities.

How to install

Follow these concrete steps to set up the MCP server locally on your machine.

Configuration and usage notes

Configure your client to connect to the MCP server using the provided local runtime setup. If you are using Claude Desktop or another MCP client, point it to the local server configuration and ensure the ChromaDB instance is running. You’ll typically initialize the system, then start building your knowledge base by storing new information and querying for related items.

Security and maintenance notes

Run everything on your own hardware to keep data private. Regularly back up your JSON exports of collections and monitor disk usage, especially when enabling watch folders and duplicate detection across large directories.

Available tools

search_context

Search for relevant information across collections with optional target collection and result limit.

store_context

Store new content with metadata into a target collection for future retrieval.

list_collections

List all available collections and their metadata.

find_similar_patterns

Find code or text patterns similar to a provided example to aid refactoring or learning.

scan_directory

Preview files in a directory before ingesting, with filters for type and extensions.

batch_ingest

Bulk ingest files into a collection with full metadata extraction.

quick_load

Rapidly load files into a temporary collection for fast processing.

unload_collection

Delete a temporary collection after quick_load usage.

export_collection

Export a collection to a JSON file for backup or transfer.

import_collection

Import a collection from a JSON file, with optional overwrite.

get_collection_info

Get detailed stats about a collection.

ingest_file

Ingest a single file into a target collection with metadata.

list_file_types

Show all supported file extensions.

extract_exif

Extract detailed EXIF metadata from photos including GPS, date, and camera data.

watch_folder

Start auto-ingesting new files from a folder into a collection.

stop_watch

Stop watching a folder for automatic ingestion.

list_watchers

List all active folder watchers.

find_duplicates

Scan directories for duplicate files and report wasted space.

compare_files

Check if two files are duplicates.

find_collection_duplicates

Find duplicate entries within a ChromaDB collection.