The official starnet github repo is androoAGI/starnet, built by Andrew Sims, and it's where you get the real StarNet desktop agent harness: the source code, the README, and the link to the prebuilt installers. This guide covers what the README says, the licence terms, how to install from releases or run it from source, how to connect models, how to import your Hermes or OpenClaw agents, and the one warning you need to know before you download anything.
📺 Watch: StarNet: FREE Pixel AI Agent Office!
🔥 Get the Agent OS as a free bonus: AI Profit Boardroom members get the full Agent OS zip, prompt libraries, daily tutorials and weekly live coaching calls. → Get inside
I showed StarNet on the Julian Goldie Podcast, and the most common question afterwards was "where do I actually get it?" Here's the clean answer.
StarNet GitHub: Where the Official Repo Lives
The official repo is github.com/androoAGI/starnet. The author is Andrew Sims, who goes by androoAGI. At the time of writing it had about 1,200 stars and 216 forks.
There's also a separate releases repo, androoAGI/starnet-releases, where the prebuilt installers live. If you just want to install and go, that's the one you'll download from.
If you want to go beyond installing tools and actually build AI agents that do work for your business, the AI Profit Boardroom has full Hermes and AI agent courses plus my Agent OS, for $69/mo.
What the README Says StarNet Is
Per the README, StarNet is "a living pixel-art station where real AI agents do real work." It's a local-first desktop agent harness. You bring your own key.
The station is a projection of live runtime state. Per the README:
- A room is a capability-scoped team.
- A hallway is an authorised handoff lane.
- A placed object is a real capability grant.
You start with one agent, then place bays or summon specialists. Each is a distinct, bounded agent run with its own workspace and permissions.
The README lists these features:
- Task briefs that turn a vague ask into one concrete question with options.
- An OUTBOX folder where deliverables land as real files.
- MCP connectors.
- Messaging through Telegram, Discord, Slack, Signal and Matrix.
- Night Shift, where agents keep working on cron schedules with guardrails.
- Push-to-talk voice.
- Spend, budgets and run history saved on disk.
One line from the README stands out: StarNet "does not simulate revenue, completed work, model activity, or spend". What you see on the station is meant to reflect what's really happening.
Is StarNet Open Source? Licence and Brand Terms
Yes, with one important split. Per the README, the code is MIT licensed. You can use it, modify it and build on it.
But the StarNet name, logo, station art and sprites are NOT licensed. So if you fork it and publish your own version, you must rebrand. You can't ship a fork called StarNet with the same art.
That detail ties straight into the warning below.
📺 Watch: Gemma 4 DESTROYS Github Copilot?
Warning: Watch Out for Forks
The repo has been forked a lot, and forks copy the same description. That means a search can turn up repos that look like StarNet but aren't the original.
Install from androoAGI/starnet and its releases repo, androoAGI/starnet-releases, only. Don't grab an installer from a random fork. An AI agent harness can get access to your files, your accounts and your API keys, so where you download it from really matters.
How to Install StarNet From Releases
This is the easiest route. Go to the androoAGI/starnet-releases repo and open the "latest" release page. Per the README, the prebuilt installers are:
| Platform | Installer |
|---|---|
| Windows 10/11 64-bit | .exe |
| macOS Apple silicon | aarch64 .dmg |
| macOS Intel | x64 .dmg |
| Linux | Not a supported public release |
Pick the right one for your machine. On a Mac, check whether you've got Apple silicon or Intel before you download. Getting that wrong is the most common install mistake.
Then run the installer and open StarNet.
Because it's local-first, everything runs on your own machine. Your agents, their workspaces, your run history and your spend records all stay on your computer. The only thing that leaves is whatever you send to the model provider you connect, and if you use Ollama, even that stays local. For a lot of people, that's the main reason to pick a desktop harness over a hosted one.
How to Run StarNet From Source
If you're on Linux, or you just want to run the code directly, you can run it from source. Per the README, you need Node.js 18 or newer, and 22 is recommended.
- Clone the repo: git clone https://github.com/androoAGI/starnet
- Go into the folder.
- Start the sidecar: node sidecar/index.js
- Open localhost:8787 in your browser.
That's it. Running from source gets you the same station in your browser instead of a packaged desktop app.
Not sure whether to go installer or source, or how this fits with the rest of your agent stack? Book a free AI strategy session and we'll work it out together.
📺 Watch: NEW Perplexity Computer Update is INSANE! (Claude Code + GitHub CLI)
Connecting Models to StarNet
StarNet doesn't come with its own brain. You connect one. Per the README, you've got three options:
- OpenRouter: paste an OpenRouter key and pick from its models.
- Provider sign-in: sign in with supported provider accounts, including Anthropic, OpenAI and Google.
- Ollama, for free: install Ollama and pick Ollama as the provider. The README's example is qwen3:8b.
Keys live in your operating system's keychain, never in the frontend. That's a good design choice for something that runs on your desktop.
When I set mine up, I signed in with Grok in one message and picked Grok 4.6, which I prefer to 4.7. You also choose a reasoning level: low, medium, high or max.
If you want to go the free local route, my guide on how to set up Hermes with Ollama walks through getting Ollama running, and the same Ollama install works as StarNet's provider. To compare which models are worth running, check Goldie Bench.
Importing Hermes and OpenClaw Agents
This is the feature Hermes users asked about most. Per the README, StarNet imports existing agents from OpenClaw or Hermes installs. It brings over:
- Persona
- Instructions
- Memory
- Model
You re-enter your API keys after import. Keys don't carry across, which is sensible from a security point of view.
So if you've already spent time shaping a Hermes agent, you don't throw that work away. You bring it onto the station and see it as a pixel character with its own room. If you like seeing Hermes agents in a visual workspace, the Hermes 3D office is another take on the same idea.
You can also plug open-source projects like StarNet into my Agent OS with the zip setup. The Agent OS guide explains it, and AIPB members get it.
A Quick Look Inside After Install
I did a full hands-on tour in my StarNet AI agent walkthrough, so here's just the short version.
On first run you name your agent, pick an appearance and a personality (warm, dry, composed or unhinged), and choose "ask for approval" or "full power". I went with ask for approval. Then you connect a brain and the agent wakes up with a little animation.
From there you get the station view, an agent settings panel, chat, widgets, an app catalog (Gmail, Notion, GitHub and more), a skill market and skill library, and a Crew menu where you recruit specialists. One tip: "deploy" overwrites your current agent's setup, while "summon" adds a second agent beside it.
My honest take: I don't think there's a massive amount of practical use cases for this yet, but it's fun, and it's the smoothest office-style agent UI I've tried.
Early-Release Caveats
StarNet is an early release. Per the README, Windows is the most-tested platform and macOS has less real-world coverage. Linux isn't a supported public release, so source is your route there.
What that means for you:
- Expect rough edges, especially on Mac.
- I noticed a slight delay on chat sends when I tested it.
- Start with approval mode, not full power, until you trust it.
- Watch your spend and budgets, which StarNet saves on disk.
- Only install from the official repo and releases repo.
If you want the step-by-step courses on Hermes, AI agents and my Agent OS, join the AI Profit Boardroom for $69/mo.
FAQ
Where is the StarNet GitHub repo?
The official StarNet repo is androoAGI/starnet on GitHub, by Andrew Sims. The prebuilt installers are on the androoAGI/starnet-releases "latest" release page. Avoid forks, which copy the same description.
Is StarNet open source?
Yes, the code is MIT licensed. Per the README, the StarNet name, logo, station art and sprites are not licensed, so any fork you publish has to be rebranded.
Does StarNet run on Linux?
There's no supported public Linux release. You can run it from source instead: clone the repo, run node sidecar/index.js with Node.js 18 or newer (22 recommended), then open localhost:8787.
Can I use StarNet for free?
Yes. StarNet is free, and you can run it with a local model by installing Ollama and picking it as the provider. Want a hand planning a free or low-cost agent stack? Book a free AI strategy session.
