MIT Open Source — Runs on your hardware

Your own AI agent,
on your own machine

Message Orbit from Discord or iMessage. It runs on your hardware with full shell, files, web, and screen access — and reports back when the work is done. Owner-locked. Private. Yours.

Discord DM
O
Orbit
npx github:GOODMAN-PRO/helm
Y
You
Build a landing page for my side project and push it to GitHub Pages
O
Orbit
On it. Building the page, adding the deployment config, and pushing now. Will screenshot the live result when done.
orbit — your-machine ~ local
# Orbit running 24/7 on your hardware
orbit@your-machine $ helm status
Status: running  |  Brain: synced  |  Model: opus
orbit@your-machine $ use windows
Switched brain to Windows workstation
Cross-platform fleet active: Mac + Windows
orbit@windows-box $ helm mind show
Memory active  47 projects  3200 recall entries
Self-improvement: scheduled tonight  |  tests passing: 142
What it does

Not a chatbot. A worker.

Orbit lives on your hardware. You message it from Discord or iMessage. It gets the work done and reports back.

Real shell access

Full shell, files, web browsing, and screen control. Not a sandboxed API call — actual execution on your machine.

Owner-locked and private

Only your Discord ID can command it. Nothing leaves hardware you own. Secrets stored in an encrypted vault.

Runs on any model

Claude subscription, an API key, or free local via Ollama. You choose what you pay — including nothing.

Mac + Windows fleet

One synced brain across all your machines. Switch between Mac and Windows with a chat command. Move files both ways.

Persistent memory

Structured recall plus Helm Mind — an AI-first second brain over a synced Markdown knowledge base that rewrites itself as you talk.

Self-improves nightly

Thinks in the background 24/7. Nightly upgrades gated by tests — auto-reverts on failure. When it gets stuck, it queues the fix for overnight.

Three steps. Then it works for you.

You message it. It acts on your machine. It reports back.

1
You message
A DM from Discord or iMessage. Describe what you need in plain language.
2
It acts on your hardware
Shell, files, web, screen — full tools running on your machine. Not a cloud sandbox.
3
It reports back
Files built, problems fixed, screenshots, summaries — straight into your Discord or iMessage.

One command. Everything else is automatic.

The installer sets up Node, Claude Code, and all dependencies. Then the setup wizard walks you through Discord, your model, and 24/7 mode.

terminal
$ npx github:GOODMAN-PRO/helm
# Works on macOS, Windows, and Linux
macOS / Linux
curl -fsSL https://raw.githubusercontent.com/GOODMAN-PRO/helm/main/install.sh | sh
Windows PowerShell
irm https://raw.githubusercontent.com/GOODMAN-PRO/helm/main/install.ps1 | iex
Manual (git)
git clone https://github.com/GOODMAN-PRO/helm.git && cd helm && npm install

How Orbit compares to OpenClaw

OpenClaw is the closest peer in the local AI agent space. Here's where Orbit's approach diverges.

Capability Orbit OpenClaw
Runs on your own hardware
Open / self-hosted MIT
Chat from Discord / iMessage
Free / local models (Ollama)
Persistent memory + synced brain Helm Mind ~
Multi-machine fleet (Mac + Windows) ~
Clicks and types in any app
Thinks 24/7 and self-upgrades nightly loop
One brain synced across fleet

Full comparison matrix (vs Hermes, Cursor, Manus, and more) at goodman-pro.github.io/helm

You bring the brain. Orbit runs the work.

No vendor lock-in. No subscription to a third party. Power Orbit with whatever you already have.

$0 / forever
Ollama — free local
Runs entirely on your machine. No API calls, no data leaves your hardware. Auto-downloads the model for you.
API key — pay per use
Anthropic / OpenAI / any
Use your existing API key. Pay only for what you use. Works with any Anthropic-compatible endpoint.
Claude — subscription
Pro / Max / Team
OAuth via the claude CLI. No per-message cost if you have a subscription. One command to connect.

The cloud has your data.
We think that is wrong.

Every cloud AI assistant you've used has a simple business model: gather your data, train on it, improve the product, sell it back to you. Your prompts, your files, your projects — all fuel for someone else's model.

Orbit changes this. It runs on hardware you own, uses a model you choose, and never sends your context to a server you don't control. Your data stays yours.

"Your AI agent, on your own machine. Private. Yours."
0
Data sent to third-party servers when running on Ollama
24/7
Background cognition with nightly self-improvement
MIT
Open-source license. Fork it, modify it, run it your way.
1
Command to install on macOS, Windows, or Linux