LLM Kosh runs on your own machine — the client starts it, so there is no endpoint to ping. 87 installs a week from pypi. Last commit 14 Sep 2026.
Local-first AI memory cartridge with persistent MCP memory for Claude via SQLite FTS5.
What the code does
We read the source, 4 d ago · rules 995398eba69f
What this server is able to do. For an MCP server this is often the job itself — a terminal server runs commands because that is what it is for. Listed so you know what you are plugging in, not as an accusation.
setup.py: выполняет код при установке пакета
this.activeDaemon = spawn(exe, fullArgs, {
return subprocess.run(
env = dict(os.environ)
Is this your server and something here is wrong? Tell us — corrections are free and do not require a plan.
This code can reach further than it looks
We found places where it runs commands, builds paths or queries from values it is given. None of that is a flaw by itself — it becomes one when the code changes, and code changes quietly between releases. We re-read it on every one.
Connect this server
This server runs on your own machine — install it with the package manager and the client starts it for you. Package name taken from the official registry entry.
claude mcp add llm-kosh -e 'CARTRIDGE_WORKSPACE=<CARTRIDGE_WORKSPACE>' -- uvx llm-kosh{
"mcpServers": {
"llm-kosh": {
"command": "uvx",
"args": [
"llm-kosh"
],
"env": {
"CARTRIDGE_WORKSPACE": "<CARTRIDGE_WORKSPACE>"
}
}
}
}[mcp_servers.llm-kosh]
command = "uvx"
args = ["llm-kosh"]
env = { CARTRIDGE_WORKSPACE = "<CARTRIDGE_WORKSPACE>" }{
"mcpServers": {
"llm-kosh": {
"command": "uvx",
"args": [
"llm-kosh"
],
"env": {
"CARTRIDGE_WORKSPACE": "<CARTRIDGE_WORKSPACE>"
}
}
}
}{
"servers": {
"llm-kosh": {
"type": "stdio",
"command": "uvx",
"args": [
"llm-kosh"
],
"env": {
"CARTRIDGE_WORKSPACE": "<CARTRIDGE_WORKSPACE>"
}
}
}
}Put your own value in place of
<CARTRIDGE_WORKSPACE> (Path to the local folder llm-kosh should use as its memory cartridge root).
The author’s registry entry marks it as required.
Alternatives to LLM Kosh
same job, measured the same wayLocal-first persistent memory for AI agents — SQLite + local embeddings, MCP-native, no cloud.
Local-first persistent session memory for Claude Code and other agents. SQLite + FTS5, no cloud.
Persistent session memory for Claude Code. Local SQLite, no LLMs, no network.
Persistent memory MCP server for Claude Code, Codex and OpenCode, stored locally in SQLite
Persistent local AI memory. SQLite + FTS5 + vector search, knowledge graph. No cloud, no API keys.
Local persistent memory for AI agents: SQLite + FTS5 (BM25) lexical search, zero external infra.
Persistent memory + decision tracking for Claude across sessions. Local SQLite, no cloud.
Persistent memory for AI coding agents. Local-first SQLite + vector search. 17 MCP tools.
LLM Kosh — questions
Answers built from our own checks of this server.