Installation

Kore Memory is distributed as a Python package on PyPI. It requires Python 3.11 or later.

Quick Install

pip install kore-memory

This installs the core library with FTS5 full-text search. It has no heavy dependencies and works immediately.

Installation Extras

Kore Memory uses optional extras to keep the base installation lightweight. Install only what you need:

Core Only (FTS5 Search)

pip install kore-memory

Includes:

• Full REST API server
• FTS5 full-text search
• Auto-importance scoring
• Ebbinghaus decay engine
• Tags, relations, timeline
• Export/import, archive
• Web dashboard

info

FTS5 search is keyword-based and works well for exact matches. For fuzzy, cross-language, and semantic similarity search, install the semantic extra.

Semantic Search

pip install kore-memory[semantic]

Adds:

• Local sentence-transformer embeddings (paraphrase-multilingual-MiniLM-L12-v2)
• Cosine similarity search across 50+ languages
• Memory compression (merges similar memories)
• No API keys or cloud calls required

tip

The default embedding model is ~120 MB and downloads automatically on first use. It runs on CPU -- no GPU required.

MCP Server

pip install kore-memory[semantic,mcp]

Adds:

• MCP (Model Context Protocol) server with 14 tools
• Compatible with Claude Desktop, Cursor, Claude Code, and other MCP clients
• Start with the kore-mcp command

All Extras

pip install kore-memory[semantic,mcp]

This is the recommended installation for most users. It gives you the full feature set.

Install from Source

Clone the repository and install in development mode:

git clone https://github.com/auriti-web-design/kore-memory.git
cd kore-memory
pip install -e ".[semantic,mcp]"

This is useful for contributing or running the latest unreleased changes.

Development Dependencies

If you plan to contribute, install the development extras as well:

pip install -e ".[semantic,mcp,dev]"

This adds testing tools (pytest, coverage), linting (ruff, mypy), and documentation utilities.

Verify Installation

After installing, verify that everything works:

# Check the CLI is available
kore --help

# Start the server
kore

# In another terminal, check health
curl http://localhost:8765/health

You should see output similar to:

{
  "status": "healthy",
  "version": "1.0.0",
  "capabilities": {
    "semantic_search": true,
    "fts5": true,
    "mcp": true
  }
}

JavaScript / TypeScript Client

If you need to connect to a running Kore Memory server from a Node.js or browser environment:

npm install kore-memory-client

Features:

• Zero dependencies
• ESM + CJS support
• Full TypeScript types
• 17 async methods
• ~6 KB minified
• Node 18+

import { KoreClient } from 'kore-memory-client';

const kore = new KoreClient({
  baseUrl: 'http://localhost:8765',
  agentId: 'my-agent'
});

await kore.save({ content: 'Hello from TypeScript', category: 'general' });
const results = await kore.search({ q: 'hello', limit: 5 });

System Requirements

Requirement	Minimum	Recommended
Python	3.11	3.12+
RAM	256 MB	512 MB+ (with semantic)
Disk	50 MB	200 MB+ (model download)
OS	Linux, macOS, Windows	Any
GPU	Not required	Not required
Network	Not required	Not required

warning

SQLite must be compiled with FTS5 support. This is included by default in Python's bundled SQLite on most platforms. If you get an FTS5 error, rebuild Python with a newer SQLite version.

Upgrading

pip install --upgrade kore-memory

Kore Memory handles database migrations automatically. Your existing memories are preserved across upgrades.

Uninstalling

pip uninstall kore-memory

Your memory database (default: data/memory.db) is not removed. Delete it manually if you want a clean slate.