Local-first AI memory cartridge: persistent, cross-session memory for Claude and any LLM via MCP
Project description
llm-kosh
Give your local LLMs a permanent, air-gapped memory cartridge.
llm-kosh is a lightning-fast, SQLite-backed memory system that gives AI assistants (like Claude and Cursor) permanent recall across sessions. Stop copy-pasting the same 15 architectural decisions and database schemas into every new chat.
By running completely locally, it guarantees zero cloud syncing, zero API costs, and absolute privacy for your proprietary codebase.
🚀 The Killer Feature: Native MCP Server
llm-kosh natively supports the Model Context Protocol (MCP). This means tools like Claude Desktop and Cursor can automatically search and read your memory cartridge without you doing anything.
Claude Desktop Setup
Simply add this to your claude_desktop_config.json:
{
"mcpServers": {
"llm-kosh": {
"command": "uvx",
"args": ["llm-kosh", "mcp", "--root", "C:/path/to/your/cartridge", "--allow-write"]
}
}
}
Now, when you ask Claude "What database are we using for auth?", it automatically searches your llm-kosh cartridge and knows the answer.
✨ Features
- 🧠 Lightning Fast Local Memory: Powered by an embedded SQLite FTS5 database. Sub-50ms search times across thousands of context files.
- ⚡ Global Quick Capture: Press
Ctrl+Shift+Spaceanywhere on your OS to instantly dump a thought, decision, or code snippet into your cartridge. - 🛡️ Air-Gapped & Secure: Built-in safety Airlock prevents destructive writes. No data ever leaves your machine unless you explicitly pack it.
- 🖥️ Beautiful Desktop UI: Comes with a glassmorphism Electron/React desktop app to manage your memory, monitor the MCP daemon, and configure hotkeys.
📦 Quickstart
Get from zero to a running memory cartridge in under 2 minutes.
1. Install the core system
pip install llm-kosh
2. Initialize a new cartridge
# Creates a new .llm-kosh root in your current directory
llm-kosh init --owner "Your Name"
3. Launch the Desktop App & Daemon
# Starts the UI, the background watcher, and the MCP bridge
llm-kosh desktop
📚 Deep Dives:
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Filter files by name, interpreter, ABI, and platform.
If you're not sure about the file name format, learn more about wheel file names.
Copy a direct link to the current filters
File details
Details for the file llm_kosh-2.1.0.tar.gz.
File metadata
- Download URL: llm_kosh-2.1.0.tar.gz
- Upload date:
- Size: 33.3 kB
- Tags: Source
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/6.2.0 CPython/3.13.13
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
9ba6a88031e5e94cec7447dca258e02d35e8e30826e464a377504948b3f90a79
|
|
| MD5 |
315e023223e893f51a6d684275caf232
|
|
| BLAKE2b-256 |
d63ee696a914b5481c22323653749ca0bbf28ae002099ce3a0eb09da56bd871a
|
File details
Details for the file llm_kosh-2.1.0-py3-none-any.whl.
File metadata
- Download URL: llm_kosh-2.1.0-py3-none-any.whl
- Upload date:
- Size: 25.6 kB
- Tags: Python 3
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/6.2.0 CPython/3.13.13
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
dfb025bb2dd3de4b2a4a682a64c32a74f3be04e5d175457e2fec8befbb147ec4
|
|
| MD5 |
6c31adac9307b7f6f93cd1939ad5debb
|
|
| BLAKE2b-256 |
b0ce2d2e10c329471ff093415706e2dcb3fb7f121144c00751f29d819c63705f
|