agentos.guide › blog
StarNet AI Agent: The Quick Start, And How It Fits Your Agent OS
StarNet AI agent is a free, open-source harness where you create real AI agents and watch them work inside a pixel-art space station on your own computer.
I tested it, signed it into Grok and had my first agent awake in one or two clicks.
I've also built it into my own Agent OS as a tab, so I'll show you exactly where it fits next to your other agents.
This is the quick start, the brains it supports, every feature I saw, and the honest limits.
What the StarNet AI agent harness is
StarNet is a local-first desktop harness for building and running real AI agent teams.
It was built by Andrew Sims, and the code is on GitHub under the androoAGI account.
The version I ran was v0.13.1, and the code is free and open source under the MIT License.
The twist is the interface, because your agents live inside a pixel-art space station.
The README says the station is a projection of live runtime state, not a decoration.
A room is a team with certain capabilities, a hallway is a handoff lane, and a placed object is a real capability grant.
The agents make real model calls, use real tools and create real costs if you connect a paid model.
The project's core rule is that the interface must never show a state the harness cannot prove.
🔥 Want the exact StarNet agent station setup I run inside my Agent OS?
Inside the AI Profit Boardroom you get the installable Agent OS, step-by-step video tutorials, four coaching calls a week and 3,400+ members building real automations.
Why I added StarNet to my Agent OS
I'll be honest with you, because that's the deal on this site.
In the video I said I think it's fun to play with, and I don't think there's a massive amount of practical use cases yet.
That's still true.
But it's the smoothest office-style agent add-on I've tested, and it reminded me of the first time I tried Hermes Agent.
It also feels like a much more fun version of Paperclip.
So I built it into my Agent OS as its own tab, right next to my other agents.
My heavy automation still runs on Hermes, Claude Code and my pipelines.
StarNet is the visual crew room where I can see agents, level them up and test ideas quickly.
StarNet AI agent quick start in seven steps
The install is short, and I've covered it in full in my StarNet GitHub guide.
You either download the desktop app for Windows or macOS, or you run it from source with Node.js 18 or newer.
On my machine the source route still needed an npm ci first, even though the README calls it zero-install.
Once it's open, the quick start looks like this.
| Step | What you do | What you see |
|---|---|---|
| 1 | Open StarNet and start the first-run screen. | It asks you to create your first agent. |
| 2 | Name the agent and choose its appearance. | A pixel-art crew member appears. |
| 3 | Pick a personality. | You choose from composed, warm, blunt, dry, unhinged or upbeat. |
| 4 | Pick ask for approval or full power. | The agent's leash is set. |
| 5 | Connect a brain by signing in or pasting a key. | Your models load into the picker. |
| 6 | Pick the model and a reasoning level. | You choose low, medium, high or max. |
| 7 | Create the agent. | A waking-up animation plays and the agent says hello. |
I signed in with Grok in one quick step and picked Grok 4.6, because I actually prefer it to Grok 4.7.
My agent woke up and said, "I'm awake. Let's give the station a direction."
That took one or two clicks after the brain was connected.
Which brain to plug into StarNet
StarNet works with sign-ins you already pay for, API keys, and free local models.
| Brain | How you connect it | Where it fits in an Agent OS |
|---|---|---|
| Grok | Sign in with your Grok account in one step. | This is what I run, on Grok 4.6, for chatty station agents. |
| Claude | Use the Claude CLI with your Claude Code subscription, or paste an Anthropic key. | It suits agents that plan and write longer pieces. |
| Kimi | Sign in with your Kimi account. | It reuses a subscription you may already pay for. |
| ChatGPT | Sign in through Codex. | It reuses a ChatGPT account you already have. |
| OpenRouter | Paste one key and pick from many models. | It is handy when you already route every model through OpenRouter. |
| Ollama | Pick it as the provider and pull a model locally. | It is free and private, but slower and rougher on long tasks. |
The README also lists API keys for OpenAI, Gemini, xAI, Groq, Mistral, DeepSeek, Together, Fireworks, Perplexity and Cerebras.
Your keys sit in the operating system keychain and never in the front end.
For the free route, the README suggests installing Ollama and pulling a model such as qwen3:8b.
Pick a model that supports tools, because StarNet agents act through tool calls.
It's honest that an 8B model needs about 10 GB of graphics memory while it works, so check your machine first.
If you want to see how these brains score on real builds, I've put the numbers on GoldieBench.
How to use every StarNet feature
Here's every feature I walked through in the video, and how each one fits an Agent OS.
| Feature | What it does | How I use it in my Agent OS |
|---|---|---|
| Station view | You zoom, change the frame and open the four docks where everything you run lives. | I keep it open as a live picture of what my crew is doing. |
| Agent detail panel | It holds the model, personality, instructions, access permissions and approval prompts. | I write a clear purpose and only grant the access the job needs. |
| Brief, Growth, History and Memories | These tabs show focus, levelling, past actions and what the agent remembered. | I check History after every run. |
| Chat and agent list | You talk to an agent and switch between idle and online agents, projects and sessions. | I use it for quick briefs to one agent at a time. |
| Widgets | The plus button adds connected apps, pinned widgets and a daily brief. | A daily brief widget gives me a morning snapshot. |
| App catalog | Add an ability with Gmail, Notion, GitHub, flights, physical letters, email and Firecrawl web search. | I only connect the apps an agent truly needs. |
| Skill market | It offers originals such as ad copy testing and calendar scheduling, with community updates. | I install skills one at a time and test each one. |
| Framework library | It holds procedures such as a decision framework, make a plan, humanising content and hard conversation. | I enable only the frameworks that match the agent's role. |
| Crew | Recruit roles like strategist, chief of staff, researcher and marketer, then Deploy or Summon. | I Summon so my first agent keeps its setup. |
| Tasks, quests and systems | Tasks, quests, a work list, builds and a Systems menu sit in one place. | I give the crew one quest at a time. |
Your agents level up the more you use them, so it feels like a Tamagotchi for an AI agent.
In chat, my agent asked, "What made you want to set up an AI agent?"
I told it SEO and gave it aiprofitboardroom.com, which is exactly the kind of first job I'd suggest.
There's a slight delay when you send a message, so give replies a moment.
The Agent OS workflow I'd run with StarNet
This is how I'd slot StarNet into a working agent system without letting it run wild.
- First, create one agent with a single clear purpose written in its instructions.
- Second, keep it on ask for approval until you've watched a few runs.
- Third, grant web access only if the job needs research, and leave local files off unless it needs them.
- Fourth, switch on one framework, such as make a plan, so the agent thinks before it acts.
- Fifth, Summon a researcher and a marketer from Crew so the first agent keeps its setup.
- Sixth, give the crew one quest or task and check the History tab when it's done.
- Seventh, move any workflow that proves itself into your main Agent OS pipeline.
That last step matters most.
StarNet is a brilliant place to prototype a crew, and your Agent OS is where the proven jobs should live.
The README features that make it more than a toy
I didn't demo all of these on camera, so I'm taking them straight from the official README.
- You can message agents from Telegram, Discord, Slack, Signal or Matrix.
- Night Shift keeps agents working while you're away, inside a leash you set, and logs every away action.
- Recipes, reusable skills and cron schedules are built in.
- Task Briefs turn a vague request into one concrete question with options.
- Finished work lands in an OUTBOX as real files.
- MCP connectors extend what agents can touch.
- Spend, budgets and run history are saved on disk and shown as they are.
Coming from Hermes Agent?
One viewer was annoyed that StarNet can't use Hermes, and that's fair.
Hermes isn't one of the brains you can pick.
The README does say StarNet can import an existing Hermes or OpenClaw agent from its home folder.
It brings across the persona, the instructions, the memory and the model it finds there.
API keys never transfer, so you re-enter those yourself.
That gives you a quick way to see your Hermes agent standing in the station.
Limits to know before you build on it
StarNet is an early release, and the README says Windows is the most-tested platform.
A paid model connected to StarNet costs real money, because every call is real.
Chat has a slight delay.
Local models are weaker on long tasks.
The MIT License covers the code, but the StarNet name, logo and artwork are not licensed for you to rebrand.
Several viewers asked for more tutorials, so expect me to keep testing it.
Also On Our Network
FAQ
What is the StarNet AI agent?
StarNet is a free, open-source, local-first desktop harness where you create AI agents and watch them work inside a pixel-art space station. The agents make real model calls and use real tools.
Is StarNet free?
Yes. The code is open source under the MIT License. The model you connect may cost money unless you reuse a subscription or run a free local model through Ollama.
Which models does StarNet support?
It supports sign-ins for Grok, Kimi, ChatGPT through Codex and the Claude CLI, API keys for OpenRouter, Anthropic, OpenAI, Gemini, xAI, DeepSeek and more, and local models through Ollama.
Can StarNet use Hermes Agent?
Hermes is not a brain option, but the README says StarNet can import an existing Hermes or OpenClaw agent, including its persona, instructions, memory and model. API keys must be entered again.
How does StarNet fit an Agent OS?
I use it as a visual crew room for prototyping agents and crews, then move any workflow that proves itself into my main Agent OS pipelines.
Final word on the StarNet AI agent
Use it to prototype a crew you can actually see, keep approvals on, and move the proven jobs into your main system.
That's how I'd put the StarNet AI agent to work inside an Agent OS.
About Julian
I'm Julian Goldie, an AI entrepreneur, SEO expert and the founder of the AI Profit Boardroom, which has 3,400+ members.
I help business owners scale with AI agents, automation and SEO.
My YouTube channel has 400,000+ subscribers, and I run Goldie Agency, a seven-figure SEO agency.
Related guides on agentos.guide
→ Grok 4.7 In The Agent OS: 20 Games, One Shot, Real Scores
→ The Goldie Living Office: Your AI Team, Made Visible
Get the whole Agent OS
The installable system behind every guide on this site — dashboard, agents, memory and pipelines — updated as new tools land, with 3,400+ members and four coaching calls a week.
Join the AI Profit Boardroom →📺 Video notes + links to the tools 👉
🎥 Learn how I make these videos 👉