Safe Context
for Agents

How to
Luthn secure shared memory architecture for AI agents: approved context, external sources, audit, and a separate encrypted vault.

WHY LUTHN

A secure shared memory layer for AI agents.

Luthn keeps approved context available across agents, while audit and an encrypted vault protect what should stay private.

SHARED MEMORY

Stop repeating yourself. Approved context stays available across agents.

EXTERNAL SOURCES

Context is scattered. Luthn gathers useful signals from messengers, documents, and mail.

AUDIT

Sharing needs a check. Audit decides what may cross into shared memory.

VAULT

Sensitive data stays out of agent memory. It is encrypted and stored separately.

[HOW TO]

Run Luthn.

Install. Check. Connect.

01.

INSTALL

Install Luthn.

curl -fsSL https://raw.githubusercontent.com/JakobSung/Luthn/main/scripts/install.sh | bash -s -- --channel stable

macOS / Linux · Docker required

02.

CHECK

Make sure it is ready.

luthn status

Open the console at http://127.0.0.1:8080

03.

CONNECT

Add agent.

luthn connect codex
luthn connection status codex

luthn connect claude
luthn connection status claude

Run after Luthn is ready. Agent connection is a separate step.

WINDOWSInstall with PowerShell
$installer = Join-Path ([IO.Path]::GetTempPath()) "luthn-install.ps1"
irm https://raw.githubusercontent.com/JakobSung/Luthn/main/scripts/install.ps1 -OutFile $installer
pwsh -NoProfile -File $installer -Channel stable
Remove-Item -LiteralPath $installer

Requires PowerShell 7.4+ and Docker Desktop in Linux container mode.