Mosesbeta in development
Windows beta · coming soon
A personal AI agent

An AI that remembers you, built to grow with you.

Moses runs on your own PC and uses your own API key. It learns from ordinary conversation: what you tell it about yourself, your work and your plans. It keeps that up to date, so every conversation starts from what it already knows.

Built to compound Runs on your PC Your own API key Code console built in
Mosesconnected
Heads up, I moved the planning meeting to Thursdays.
you · Monday
Got it. Planning meetings are on Thursdays now, not Tuesdays. I'll use that from here on.
your agent
What's on for this week?
you · Wednesday, new conversation
Tomorrow is your planning meeting, the first one on the new Thursday slot.
memoryPlanning meetings moved to Thursdays (you, Monday) · replaces "Tuesdays"
your agent
Message your agent…
Send

What makes it different

Learns as you talk
No forms, no setup
You just talk. Moses picks out the facts that matter about you, your work and your plans, and saves them as memories it can use later.
Stays current
New facts replace old ones
When something changes, the newer fact takes over from the old one, and your own words count for more than a document about you.
No restarts
Memory takes effect right away
What you say can come back later in the same conversation and in the next one. It also organises its memory in the background, so you never have to.
Yours
Your PC, your key, your memory
Everything is stored on your own computer. You choose the AI model and use your own API key (from Anthropic or OpenRouter). You can see, edit, export or delete any memory.

Built for doing things, not just chatting

Moses can write and run code, work with your files and search the web. You can see everything it runs in the built-in console, and it asks you before it changes anything.

You choose the model: Claude, or an open model through OpenRouter. Each one is tested before we call it supported.

# your agent, in the built-in console
> reading budget_2026.xlsx
> writing summarize.py
? run summarize.py in Documents\Budget — allow?
✓ allowed · done in 1.8s

How the memory works

01 · WRITE
Every turn
It picks out what matters
After each message, a memory writer pulls out the facts worth keeping (who you are, what you're working on, what you've decided) and saves each one as its own memory, linked to what you said and when.
02 · UPDATE
When things change
Newer facts take over
When a new fact contradicts an older one, the newer one replaces it, and the old one is kept on record instead of being silently mixed in. Your own words count for more than a document about you.
03 · RECALL
Before it answers
It finds the right memories
Each time you write, it searches its memory by meaning and by keywords, and adds the most relevant memories to what the model sees, along with a short profile of you that's always included.
04 · ORGANISE
In the background
It tidies up while you're away
Background jobs rebuild your profile and organise memories, on a schedule and when the app starts. If your PC was off overnight, it catches up the next time it starts. Nothing needs a restart.

Under the hood, memories are stored in a local database on your PC, and the search model that understands meaning also runs locally. Only the conversation itself is sent to the AI model you choose.

Your data, your control

  • Stored on your PC. Your memories, conversations and settings live in a folder on your own computer, not on our servers.
  • Sent only to your AI provider. To answer you, your message and the relevant memories go to the model you picked (Anthropic or OpenRouter), under their terms.
  • See and change everything. View, edit, export or delete any memory, or all of them at once.
  • Your key, kept safe. Your API key is stored with Windows' own secure credential storage.
  • No tracking. Moses doesn't send us usage data.
Before it acts
You stay in charge of your computer
Moses asks before it writes files, runs code or uses the internet, and it only works inside the folders you choose. You can allow something once, or always for a particular folder.

The beta

  1. Download one installer for Windows. No Python, no terminal.
  2. Add your own API key and pick a model.
  3. Name your agent. It's yours, so call it whatever you like.
  4. Start talking. It remembers.
Status
In development
We're building the Windows beta now and will open it to a small group of testers first. Mac support comes after the beta. Learn more at getmosesai.com.
Who it's for
Developers and everyone else
You don't need to code. If you can install an app and copy an API key, you can use Moses. Developers get the built-in console as well.
What testers do
Use it, and tell us what happens
Use it in your everyday work, and report anything that breaks or surprises you with the "report a problem" button in the app.
How it opens
Small first, then wider
A small group of hand-picked testers goes first. Then we open more places. Follow along for updates.

FAQ

Do I need to know how to code?

No. The beta installs with a single Windows installer, with no Python and no terminal. You add your API key, name your agent and start talking.

Does my data leave my PC?

Your memories and settings stay on your PC. To answer you, your message and the memories relevant to it are sent to the AI provider you chose, under that provider's terms.

What does it cost?

You pay your AI provider directly for what you use, with your own API key. Moses shows you what each conversation costs, and you can set a daily spending limit.

Which AI models does it work with?

Claude (with an Anthropic API key) and an open model through OpenRouter. We test every model on our memory test before we list it as supported.

What is an API key?

It's a private code from an AI provider such as Anthropic or OpenRouter that lets Moses use their models on your account. You create one on the provider's website and paste it into Moses once.

Does it work on Mac or my phone?

The beta is Windows only. Mac comes after the beta. Later, you'll be able to reach your agent from your phone's browser, while your memory stays on your PC.

Can I give my agent its own name?

Yes. The product is Moses, but your agent is yours, so you name it the first time you run it.

When is the beta?

It's in development now. We'll share a date when we have one we can keep.

Contact

General
Say hello
general@mosesai.ai
Help
Support and beta feedback
support@mosesai.ai
Privacy
Your data requests
privacy@mosesai.ai
Updates
Follow the build on X
@DMMelzer