StarNet Reviews: My Verdict, The Setup, And The Brains That Make It Work

StarNet reviews keep landing on the same question, and my answer is that this free, open-source AI agent harness earns 3.5 out of 5.

StarNet is a local-first desktop app where real AI agents work inside a pixel-art space station on your own computer.

I'm Julian Goldie, and in the video above I set it up, create an agent and walk through the whole station.

It's the smoothest and most fun office-style agent tool I've tested, and I don't think there's a massive amount of practical use cases yet.

This page gives you my verdict and scores first, then the setup and features, and then a late section on which brains to plug into it.

If you want my team to look at how your business can get found in Google and AI search, you can book a free SEO strategy session.

My StarNet verdict and scores

My overall verdict is 3.5 out of 5.

Every score is my own opinion from my own install of v0.13.1, so treat it as one tester's view.

These scores are about StarNet as a product, not about any model, and they aren't GoldieBench scores.

What StarNet is and who makes it

StarNet is built by Andrew Sims, who publishes it on GitHub as androoAGI/starnet.

The repo calls it "a living pixel-art station where real AI agents do real work."

It's open source under the MIT License, and the repo has more than 1,500 GitHub stars.

The MIT License covers the code only, because the StarNet name, logo and artwork belong to the creator.

The README says agents make real model calls, use real tools and create real costs, and the interface must never show a state the harness can't prove.

I tested v0.13.1, and the latest release on 16 October 2026 is v0.13.5.

People also search for it as "Startnet OS", and my Startnet OS reviews post explains that name.

How to set up StarNet

The easy route is the desktop app for Windows or macOS from the StarNet releases page.

The source route needs Node.js 18 or newer and Git.

The README says the sidecar runs on Node core modules only, but my source install still needed npm ci before it would run.

The README opens the app on port 8787, and mine runs on port 8788.

The README says Windows is the most-tested platform and macOS has less real-world coverage.

Creating the first agent is quick.

You name it, check its look, pick a personality such as warm, dry, composed or unhinged, and choose "ask for approval" or "full power".

Then you connect a brain, pick the model and choose a reasoning level from low, medium, high or max.

The agent is made in one or two clicks, and mine woke up saying, "I'm awake. Let's give the station a direction."

If you'd like my team to map out where AI agents fit your marketing, get a free SEO strategy session with us.

Every StarNet feature I tested

The station view lets you zoom in and out, change the frame and go to a full zoom.

Everything you run lives in four docks, which you can open, show all at once or drag around.

Agents level up the more you use them, so it feels like a Tamagotchi for an AI agent.

The agent detail panel shows the model, personality, instructions and purpose, access permissions and approval prompts.

You can allow or remove web access and local file access, and you can change the personality any time.

The Brief, Growth, History and Memories tabs show what an agent is for, how it's grown, what it did and what it remembered.

Chat shows each agent's online status and level, and the left-hand list switches between idle and online agents.

You can manage projects and sessions there as well.

My agent asked, "What made you want to set up an AI agent?", and I told it SEO and gave it aiprofitboardroom.com.

There's a slight delay when you send a message.

The plus button adds widgets, including connected apps, pinned widgets, an app chooser and a daily brief.

The app catalog, labelled "add an ability", includes Gmail, Notion and GitHub, and most apps sign in directly while some need an API key.

It also has real-world categories such as finding flights, sending physical letters, email and web search through Firecrawl.

The skill market has standard originals such as ad copy testing and calendar scheduling.

The skill library holds frameworks such as a decision framework for hard technical problems, making a plan, humanising content and handling a hard conversation, and each one can be switched on or off.

Crew and Recruit give you roles like strategist, chief of staff, researcher and marketer, each with a name and appearance.

Deploy overwrites an existing agent's setup with that role, and Summon adds a new agent next to yours.

Tasks, quests, a work list, builds and a Systems menu all sit in one system.

StarNet felt like a much more fun version of Paperclip to me, and it reminded me of the first time I tested Hermes Agent.

Limits and what viewers asked

My honest line from the video was, "I just think it's fun to play with. I don't think there's a massive amount of practical use cases."

Chat has a slight delay, and the README calls the app an early release.

Several viewers asked for more tutorials.

One viewer was frustrated that they couldn't use it with Hermes Agent.

The README doesn't list Hermes Agent as a brain, but it says StarNet can import an existing Hermes or OpenClaw agent's persona, instructions, memory and model, with API keys re-entered.

I haven't re-tested since v0.13.1, and the v0.13.5 release notes say your lead agent can now add crew to a chat when you ask and that Claude Code runs are faster and cheaper.

I run StarNet as a tab inside my own Agent OS, which is my own product inside the AI Profit Boardroom, rather than as my main tool.

If you want my team to look at your site, your competitors and the keywords you're closest to winning, grab a free one-to-one strategy session.

StarNet pros and cons

What the v0.13.5 release notes add

I tested v0.13.1, so everything in this section comes from the official release notes rather than my own test.

The notes say you can bring crew into a conversation just by asking your lead agent to add them.

They say work your lead hands off gets its own threads and one live card that shows who is doing what.

They add a NEEDS YOU list that gathers every approval and question waiting on you in one place.

They say building conveyor lines is now drag-and-drop, and a finished job on one line can be sent to another line.

They say a skill an agent creates or changes now waits for a passing evaluation and your approval before it's used.

They also tell you to save a station backup first if you might ever roll back to 0.13.3 or older.

If those changes work as described, the chat and practical-use scores could go up when I re-test.

Who should use StarNet

StarNet is right for operators and creators who want to learn how agent teams behave by watching one work.

It's right for anyone who wants a free, local-first harness with clear permissions and a strong crew system.

It's not right yet for anyone who needs hardened, client-facing automation today.

If that's you, run your real jobs on a proven system and keep StarNet as the fun side project.

Does StarNet ship its own model?

No, StarNet doesn't ship its own model.

The brain you connect decides how smart your agents are, so this is where GoldieBench helps.

In my video I explained that StarNet can plug into a Claude Code subscription through the Claude CLI, Kimi and Grok.

The README adds an OpenRouter key and sign-ins for supported Anthropic, OpenAI and Google accounts.

GoldieBench scores one-shot builds out of 10, so it measures how well a model builds, not how well it chats inside a station.

Check StarNet's brain screen for the exact models on offer before you pick one.

Grok 4.6 vs Grok 4.7 for StarNet

I ran my StarNet agents on Grok 4.6, because I actually prefer it to Grok 4.7.

On GoldieBench, Grok 4.7 averages 7.15 and Grok 4.6 averages 5.95, each across 20 scored tasks.

Both live model pages list the same price of $2 in and $6 out per million tokens.

That gap matters if your StarNet agents build pages or apps, so I'd switch a builder agent to Grok 4.7.

For a chatty station agent, my own preference for Grok 4.6 still stands, because builds aren't what that agent does.

Which brain I'd pick for each StarNet role

These are my suggestions, not tested StarNet results, so try two brains before you settle.

For the free local route, the README's example model is qwen3:8b through Ollama, and that model isn't on GoldieBench.

My local models leaderboard shows the on-device models I have scored.

My StarNet AI agent post covers the first-run walkthrough in more detail.

If you'd like my team to build your plan for getting found and recommended in Google and AI search, book a free SEO strategy session.

Also On Our Network

• Agent Operatorsmy hands-on StarNet setup review
• AI Income Deskwhether StarNet is worth it for a business
• AI Tool Verdictthe full StarNet verdict with pros and cons
• agentos.guidewhy I run StarNet as a tab in my Agent OS
• aiprofitboardroom.commy StarNet AI agent walkthrough on the AI Profit Boardroom blog

If you want a clear plan for getting your content found in Google and AI search, book your free strategy session with my team.

Pick a strong brain, start every agent on "ask for approval", and you'll get the best out of what these StarNet reviews describe.