DEV Community

Wael Samoum
Wael Samoum

Posted on

I got tired of re-explaining my stack to every AI agent. So I built Auxly-Memory.

The Problem

Every AI tool I use keeps its own memory in its own silo.

I tell Claude my stack. I open Cursor — it knows nothing.
I switch to Gemini — start over from scratch.
I use Codex — explain everything again.

I was spending 10-15 minutes every single day just re-establishing
context with my own tools. That's not a workflow. That's a tax.

What I Built

Auxly-Memory is a local-first CLI that gives every AI agent on
your machine one shared memory vault.

Every agent — Claude, Codex, Gemini, Cursor, Copilot, Warp —
reads from and writes to the same folder of plain Markdown files
on your machine.

Teach one agent something once. Every other agent knows it instantly.

How It Works

Auxly-Memory is a single static Go binary that plays three roles:

1. MCP Server
Exposes memory tools (read, write, search, sync) to any
MCP-capable agent over stdio. Agents call them like any other tool.

2. Trust Gate
Every write is checked against the writing agent's trust level:

  • auto — writes land immediately
  • require_approval — queues for your review
  • read_only — can read but never write

3. Memory Vault
Accepted writes land as Markdown in ~/.auxly/memory/,
auto-committed to Git if you want.
~/.auxly/memory/
├── identity.md
├── preferences.md
├── infra.md
├── projects.md
├── products.md
└── .audit.log

What Makes It Different

Most "AI memory" tools require:

  • A cloud account
  • An API key to their service
  • Embeddings and vector databases
  • Docker and background daemons
  • Trusting a third party with your context

Auxly-Memory requires none of that.

Your memory is just Markdown files on your machine.
You own them. You can read them, edit them, grep them,
version them in Git. Nothing is locked inside a vendor's cloud.

Install

macOS / Linux:

curl -fsSL https://auxly.io/cli | sh
Enter fullscreen mode Exit fullscreen mode

Windows:

irm https://auxly.io/cli.ps1 | iex
Enter fullscreen mode Exit fullscreen mode

Homebrew:

brew install Tzamun-Arabia-IT-Co/homebrew-tap/auxly
Enter fullscreen mode Exit fullscreen mode

Quick Start

# First run — setup wizard
auxly

# Wire every agent on your machine to the shared memory
auxly setup

# Inside any connected agent's chat
/auxly-sync I use Go and deploy on Linux servers
/auxly-memory  # shows what every agent now knows about you
Enter fullscreen mode Exit fullscreen mode

The Result

Every agent you open already knows:

  • Your stack
  • Your preferences
  • Your active projects
  • Your infrastructure
  • Anything you've taught any other agent

No re-explaining. No lost context. No vendor lock-in.

Links

Would love feedback from anyone working with MCP or
juggling multiple AI agents daily.

Top comments (0)