StarNet AI agent is a free, open-source agent harness that puts real AI agents inside a pixel-art space station, and my verdict is 3.5 out of 5.

It's the smoothest and most fun office-style agent tool I've tested.

It also isn't yet the tool I'd run serious client work on, and I'll show you exactly why.

This review explains what StarNet is, how to set it up, and then scores every feature I tested one by one.

At the end you'll get a scores table, the limits that matter, and who I think should use it.

The Short Verdict On StarNet AI Agent

StarNet earns 3.5 out of 5 from me.

It wins on fun, setup speed, visibility and the crew system.

It loses points on everyday practical value and on early-release rough edges.

If you want to learn how agent teams work in a way you'll actually enjoy, it's an easy yes.

If you need a hardened system for client delivery today, I'd look elsewhere first.

What StarNet Is And Who Made It

StarNet is a local-first desktop harness for creating and running real AI agent teams.

It's built by Andrew Sims, and the code is published on GitHub under the androoAGI account.

The version I reviewed was v0.13.1.

The code is open source under the MIT License, which means you can use, modify and redistribute it, including commercially.

The brand is a separate matter, because the StarNet name, logo and artwork are not licensed with the code.

The core idea is a pixel-art space station where every agent you create has a body, a level and a place to work.

The README is strict that the station isn't a simulation.

It says agents make real model calls, use real tools and create real costs.

It also sets a rule that the interface must never show a state the harness cannot prove.

I like that rule a lot, because fake dashboards are one of my biggest frustrations with agent tools.

How To Set Up StarNet Before You Judge It

Setup is short.

You download the desktop app for Windows or macOS, or you run it from source with Node.js 18 or newer and Git.

On my machine the source route still needed an npm ci before it ran, even though the README calls it zero-install.

My full walkthrough of the repo, requirements, licence and updates is in my StarNet GitHub review and setup guide.

Once it's running, the first screen asks you to create an agent, and that's where the review really starts.

StarNet Features Scored One By One

Every score below is my own judgement from testing, out of 10.

These aren't lab benchmarks, so treat them as one experienced operator's opinion.

Agent creation: 9 out of 10

You give the agent a name, look at its pixel-art appearance and change it if you like.

You pick a personality, and the source code lists six of them: composed, warm, blunt, dry, unhinged and upbeat.

You choose "ask for approval" or "full power", then connect a brain and pick a model and reasoning level.

The whole thing takes one or two clicks, then a waking-up animation plays.

My agent said, "I'm awake. Let's give the station a direction."

It loses a point only because a beginner might not know which personality or reasoning level to pick.

Brain and model options: 8 out of 10

You can sign in with an account or paste an API key.

StarNet supports sign-ins for Grok, Kimi, ChatGPT through Codex and the Claude CLI, which uses a Claude Code subscription.

It also takes API keys for OpenRouter, Anthropic, OpenAI, Gemini, xAI, Groq, Mistral, DeepSeek, Together, Fireworks, Perplexity and Cerebras.

For free use, you can pick Ollama and run a local model.

Reasoning levels run from low through medium and high up to max.

I signed in with Grok in one step and picked Grok 4.6, because I actually prefer it to Grok 4.7.

It loses points because local models are weaker for this, which the README openly admits.

Station view: 8 out of 10

You can zoom in and out, change the frame and go to a full zoom.

Everything you run lives in four docks that you click to open, show all at once or drag around.

It's clean and readable, although it takes a few minutes to learn where everything lives.

Levelling and growth: 9 out of 10

Your agents level up the more you use them.

It turns using AI agents into a game, a bit like a Tamagotchi for an AI agent.

That's a gimmick, but it's a gimmick that genuinely makes you want to come back.

Agent settings and permissions: 9 out of 10

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

You can allow or remove access to the web and your local files for each agent.

The panel has Brief, Growth, History and Memories tabs, so you can see what an agent did and what it remembered.

This is one of the clearest permission setups I've seen in any agent harness.

Chat and agent list: 6 out of 10

Chat shows each agent's online status and level.

The left-hand list lets you switch between all agents and see which are idle and which are online.

You can manage projects and sessions here too.

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

It loses points because there's a slight delay when you send a message.

Widgets: 7 out of 10

The plus button at the top adds widgets.

You get connected apps, pinned widgets, a chooser for apps and a daily brief, and you choose what each one shows.

It's useful, although it's not the reason you'd pick StarNet.

App catalog: 8 out of 10

The app catalog is labelled "add an ability".

It includes Gmail, Notion and GitHub, and most of them 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 real-world abilities are the most surprising part of the whole tool.

Skill market and framework library: 8 out of 10

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

Some skills come built in, some install later, and the community can push updates.

The skill library holds frameworks, which are the procedures the AI follows.

Examples include a decision framework for hard technical problems, making a plan, humanising content and preparing for a hard conversation.

You can enable or disable each framework, which is a nice level of control.

Crew recruit, deploy and summon: 9 out of 10

Crew and then Recruit gives you roles like strategist, chief of staff, researcher and marketer.

You give each one a name and an appearance.

Deploy overwrites an existing agent's setup with that role.

Summon adds a new agent that appears standing next to your others in the station.

This is where StarNet feels most like a real team.

Tasks, quests and systems: 7 out of 10

Tasks, quests, a work list, a place to build new stuff, a Systems menu and builds all sit in one system.

It's ambitious, and it takes time to understand how the pieces fit together.

StarNet AI Agent Scores At A Glance

Feature My score out of 10 Why it scored that way
Agent creation 9 It takes one or two clicks and feels polished.
Brain and model options 8 Many sign-ins and keys are supported, and local models are weaker.
Station view 8 It's clear once you learn the four docks.
Levelling and growth 9 It makes agent use feel like a game.
Settings and permissions 9 Web and file access are easy to allow or remove.
Chat and agent list 6 It works well, but replies have a slight delay.
Widgets 7 They're useful but not the main draw.
App catalog 8 Real-world abilities like flights and letters stand out.
Skills and frameworks 8 Frameworks can be switched on and off per agent.
Crew recruit, deploy and summon 9 It turns one agent into a visible team.
Tasks, quests and systems 7 It's ambitious but takes time to learn.

🔥 Want the agent tools I actually rate higher for business work? Inside the AI Profit Boardroom, I've got full courses on Hermes Agent, Grok Bot, DeepSeek Harness, Claude Opus 5.5 and Jev AI. You'll be alongside 3,400+ members, with four coaching calls a week. → Get access here

Features The README Adds Beyond My Test

I didn't test all of these on camera, so I'm reporting them from the official README.

  • You can wire agents to Telegram, Discord, Slack, Signal or Matrix.
  • Night Shift lets agents keep working while you're away, inside a leash you set, with every away action logged.
  • 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 let you extend what agents can touch.
  • Push-to-talk voice is supported.
  • Spend, budgets and run history are saved on disk and shown as they are.

Where StarNet Falls Short

My first honest point is the one I made in the video.

I said, "I just think it's fun to play with. I don't think there's a massive amount of practical use cases."

That's the main reason it doesn't score higher.

The second point is the slight chat delay.

The third point is maturity, because the README calls it an early release and says Windows is the most-tested platform.

The fourth point is Hermes.

A viewer was annoyed that StarNet can't use Hermes Agent, and that's fair because Hermes isn't a brain option.

The README does say StarNet can import an existing Hermes or OpenClaw agent, bringing across the persona, instructions, memory and model, while API keys must be re-entered.

The fifth point is cost awareness, because a paid model connected to StarNet runs up a real bill.

StarNet Vs Paperclip And Hermes Agent

StarNet felt like a much more fun version of Paperclip when I tested it.

It also reminded me of the first time I tested Hermes Agent, which is high praise from me.

Question StarNet Paperclip Hermes Agent
What is the main appeal? It's a visual, gamified agent station. It's an agent team organiser. It's an always-on agent that runs real jobs.
How does it feel to set up? It's the smoothest office-style setup I've tested. It feels less playful in my experience. It takes more setup but rewards it.
What would I use it for? I'd use it for learning agent teams and light jobs. I'd use it for coordinating agents. I'd use it for serious daily automation.

For more verdicts in the same style, see my DeepSeek Harness review.

If you're choosing a brain to plug into StarNet, my list of the best AI models for coding will help.

Who Should Use StarNet

StarNet is right for curious operators, creators and small teams who want to see agent teams in action.

It's right for anyone who learns better when software feels like a game.

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

I run it as a tab inside my own Agent OS, next to my other agents, rather than as my main tool.

🔥 Want StarNet built into a full agent operating system? The AI Profit Boardroom has the agentic operating system where you can build StarNet in, with step-by-step tutorials and weekly coaching calls. → Join the AI Profit Boardroom

If you want a second opinion on which AI tools fit your business, you can book a free strategy session with my team.

Related Reading

Also On Our Network

FAQ: StarNet AI Agent Review

Is the StarNet AI agent any good?

It's very good at making a team of real AI agents visible and fun to run.

I rate it 3.5 out of 5 because it's an early release and I don't see many practical business use cases yet.

Is StarNet free and open source?

Yes, the code is open source under the MIT License.

The StarNet name, logo and artwork are owned by the creator and aren't covered by that licence.

Is StarNet better than Paperclip?

For fun and ease of setup, I think it is, because it felt like a much more fun version of Paperclip.

For serious multi-agent work, I'd judge each tool on the specific job.

Does StarNet work with local models?

Yes, you can pick Ollama and run a model on your own machine for free.

The README warns local models are slower and rougher on long tasks, and an 8B model needs about 10 GB of graphics memory.

What is the weakest part of StarNet?

For me it's the slight chat delay and the early-release rough edges.

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

📺 Video notes + links to the tools 👉

🎥 Learn how I make these videos 👉

🆓 Get a FREE AI Course + Community + 1,000 AI Agents 👉

About Julian

I'm Julian Goldie, an SEO entrepreneur, author and founder of the AI Profit Boardroom, which has 3,400+ members.

I help business owners scale with AI agents, automation and SEO.

  • I've built a 7-figure agency, Goldie Agency, with a team of around 50 people.
  • I've grown a YouTube channel to 400,000+ subscribers.
  • I wrote the Amazon best-sellers "SEO Link Building Mastery" and "Agency Marketing Mastery".
  • My Udemy courses have taught over 50,000 students.

→ Get my best AI training inside the AI Profit Boardroom

Fun, visual and free, with real rough edges, is my final verdict on the StarNet AI agent.