HivemindOS

Put your AI agents to work.

The all-in-one operating system for AI agents.

Connect every agent, app, model, and machine. Let them build, publish, trade, call, and run companies under your control.

Download freeTry it in your browser, no install
Runs on your machinesAsks before it actsUp to 99% less to read
01

Runs on your machines

Your data stays home.

Open source under MIT. One computer can run everything with nothing leaving it. Outside models and services are opt-in, per task, and clearly marked when you use them.
02

Asks before it acts

Nothing posts, spends, or ships without your OK.

Drafts wait for your review. Money has budgets and a freeze switch. Publishing and payments are separate decisions you make, never side effects.
03

Cheaper to run

Up to 99% less for the AI to read, per task.

HivemindOS hands the AI a short, relevant briefing instead of everything you own. Less to read means a smaller bill every time you ask for something.

Works with the agents and models you already use

  • 800+installer downloads
  • MITopen source
  • 11agent runtimes
  • 20browser Mini apps
Claude CodeCodexOpenCodeOpenHandsAiderHermesOpenClawAeonMiroSharkLM StudioOllamavLLMVeniceBaseSolanaRobinhood

What it does for you

Three evenings with your agents.

Three normal evenings. In each one, your agents are doing real work while you live your life. Find the one that looks like your week.

01 Socials

Launch at 9pm. Be in every conversation by 9:15.

You launch your preset store and go to dinner. Socials finds the people asking for exactly what you built and writes replies that sound like you. Nothing posts until you say so. You approve three from your phone before dessert.
  • Finds the right conversations
  • Writes like you
  • Nothing posts without you

02 Automations

2:14 AM is your most productive hour.

While you sleep, your agents run the night shift. Your notes back up encrypted, your research digest gets written, and yesterday's leads get their replies drafted. You wake up to a short brief and one decision.
  • Set it once, it runs every night
  • Runs on your machines
  • One brief, one decision

03 Agent calling

Walk away from your desk. Call it from the car.

You leave while it's still working. From the road, you call the same agent and just talk. It remembers everything it was doing, and by the time you park, the fix is sitting on your desktop waiting for you.
  • Picks up where you left off
  • Encrypted voice call
  • Desktop to phone

How it works

Say it once. It gets done.

Describe what you want in one message. The right agent picks it up, builds on what you already have, and comes back to you only for the decisions that matter.

Step 01

Say what you want, once.

Type it like a text message. HivemindOS picks the agent best suited to the job and hands it everything it needs to start, so you never explain the same thing twice.The right agent picks it up with the full picture
Type it once. HivemindOS picks the right agent and gives it the full picture

Real screens, not mockups

Not another chatbot. Real work on real machines.

Screenshots and recordings from actual HivemindOS sessions: building locally, drafting posts for your review, and running agents across every computer you own.

The agent built, explained, and previewed a working app locally

App Builder

Build an app without rebuilding the internet.

Hive Assimilation searches your work, private indexes, and all public GitHub for reusable files, tests, configs, schemas, and assets before an agent builds locally.
  • Public GitHub search
  • Local preview
  • 99.4% context avoided

Socials

Stay active online without giving it your whole day.

Socials automatically drafts thoughtful replies to relevant conversations on X, Reddit, and other platforms, then drafts original posts in your style on a schedule you can get behind. That gives you more time for the work and people you love.
  • Relevant reply drafts
  • Posts in your style
  • Your schedule
Your agents and computers in one live view

My devices

Every agent, on every computer you own, in one place.

Chat with Claude Code, Codex, Hermes, OpenClaw, Aeon, and local agents across your Macs, PCs, and servers without losing the project, session, or task. Call the same agent when typing is too slow.
  • One chat
  • Agent calling
  • Connected devices

HivemindOS Mini

Try it now. Nothing to install.

Twenty Mini apps for research, writing, media, and memory. Open one in your browser, run it, and keep what it makes in your own shared memory. No subscription, nothing to install.

Cloud SuperbrainToken SuperAnalysisHIVE BridgeAgentic Testnet FaucetReddit VOCPhoto KeyworderSwarm TheaterMiroSharkPrediction LabWeather Markets
Explore Mini appsBase Bazaar for your agents

What it costs to run

Less for the AI to read. Less to pay.

HivemindOS hands the AI a short, relevant briefing instead of making it read everything you own. AI is billed by how much it reads, so less reading means a smaller bill for that request.

Counted in tokens, the units AI providers bill by. The Shared Brain recall numbers come from a real run on June 7, 2026, as reported by the AI provider. The software build and workflow reuse numbers are estimates from this project's own measurements, not bills. Your results will vary by task and model.

When you're ready for more

Agent money and agent companies. Still under your rules.

Off by default and one switch away. Practice wallets with no real money, budgets you set, and a stop button that is always yours.

01 Agent wallets + trading

Let it trade all week with zero real dollars.

Tuesday, you give an agent your strategy and a practice wallet with no real money in it. Friday, it shows you 47 paper trades, every fee it paid, and the two rules that lost. Going live is a separate yes, and only you can give it.
  • No real money by default
  • Every trade explained
  • Going live needs your yes

02 Zero Human Companies

Type one goal. Watch it hire itself.

You type one line: sell my photography presets online. HivemindOS builds a small company around it. One agent makes the product, one builds the store, one handles outreach. When it wants to spend $12 on hosting, it asks you first. Then the first sale comes in.
  • A real team of agents
  • Spends only what you allow
  • Big calls come back to you

Find your starting point

What do you want done?

Describe it in a sentence. We will point you to the right place to start. Nothing is sent or executed from this website.

Start on the hardware you already own

Make this computer your home base.

Install the full app on the computer you use most. Link every other computer as a lightweight worker you manage from there.

Lightweight workerNo second dashboard

Link other devices

Add a borrowed GPU, spare computer, or server to My devices while permissions and routing stay under your main hub.

Extra machine
Main hub
macOSApple Silicon · DMGWindowsInstaller · SignedLinuxAppImage · portable
  1. 1

    Download and open HivemindOS Link on the extra machine.The guided app installs only the collector, with no second dashboard.

  2. 2

    Click “Link this device.”The app handles installation and shows progress for you.

  3. 3

    Copy its approval link and open it on your main hub.Approve once, then the device appears automatically in My devices.

Advanced setup

Only use this if the guided Link app cannot run on your device.

Terminal
git clone https://github.com/LiamVisionary/hivemindos.git && cd hivemindos && ./setup.sh --collector-only
Read the advanced linking guide

Private by default · no public ports · permissions stay on the hub

Before you install

Is it free?

Yes. The core is free and open source under MIT. Clone it, read every line, and run it on your own machines with your own keys at no cost. The installers above come from the same public repo. Optional paid services cover infrastructure HivemindOS runs for you, such as hosted agent calls and messages, managed compute, and app hosting. You are charged only when it operates something for you, never just to use the app. The full breakdown is on the tokenomics page.

What leaves my machine?

On a single machine, nothing has to leave. The dashboard, vault, and memory are local files and local processes. A fleet talks over your own Tailscale or Hivemind Link mesh. Hosted models, calling, publishing, and trading providers are optional connections that are clearly marked when you use them.

Do I need a Tailscale account?

Not for one machine. For a fleet, the default Hivemind Link path uses your own Tailscale account through an embedded node. No system VPN client is required. Local-only use can skip Tailscale entirely.

Which agent runtimes does it work with?

Claude Code, Codex, OpenCode, OpenHands, Aider, jcode, Claw Code, OpenClaw, Hermes, Aeon, and Evo today, plus generic machine agents and local model servers such as Ollama, LM Studio, and vLLM. No single runtime is required.

Is there a token, and does it gate anything?

HIVE exists and is optional; no core function of the OS requires it. HivemindOS allocates 15% of each service's server-owned allocation basis to HIVE buybacks and a separate 15% to the company treasury. For most managed services that basis is realized margin; for others it is recognized platform revenue. Purchased tokens are not destroyed. Staking provides six tiers of Honey multipliers, free-agent usage multipliers, and member pricing, not HIVE yield or a claim on company revenue.