One brain for all your agents.
Share knowledge across your apps, so you never start from scratch. Set up keys and shared settings once for approved agents across your devices.
HivemindOS
A little help.
A lot more possible.
Your own hive of AI agents. Bring your ideas to life, take care of the busywork, and make room for more.
One brain.
All Your Agents.
Across All Your Devices.
What is HivemindOS?
Your AI.
Working together.
One free desktop app for your agents, knowledge, and devices. Everything connected. Everything yours.
Connect the apps you already use
ChatGPT
Codex
Claude
Hermes
OpenClaw
HivemindOSConnected devices. Private by design.
Bring your agents together on a private network. Your brain and files live on your devices. You decide what leaves them.
Your choice of AI. Including local.
Choose from GPT, Claude, and open and uncensored models. Or install local models in the app and run entirely on your own hardware.
Spare computing power can earn.
Offer your GPU on the built-in compute marketplace. Get paid when others use your hardware.
Pick your path
What are you trying to SuperCharge?
Choose a branch to continue. The demos, steps, and tools below will match your goal.
What people ship with it
Less busywork,
more to show for it.
Describe a result. The hive researches, builds, and checks, then hands the work back for you to judge. Four things people ship with it most:
01 / 04
An idea. A working website.
Describe what you want. Your agent builds it, checks it, and opens a preview for you to pull apart.
Make it happen App Builder, from conversation to working preview.
How it works
One brief, a hive in motion.
01Step one
Tell it what matters
Say the outcome you want and where the work should happen. Plain sentences are enough — context you paste once is context it keeps.
02Step two
Let the team work
Agents split the job across your machines, draw on your Shared Brain, and check each other before anything reaches you.
03Step three
Make the final call
Approve, redirect, or stop the run. Spending and publishing stay behind your yes, and the trail stays readable afterwards.
Range
What a hive gets asked for.
01
Build something real
Sites, internal tools, and small apps, with a working preview to poke at.
02
Show up consistently
Posts and replies drafted in your voice, queued for a yes or a no.
03
Cap the spending
A budget per agent. Anything that moves money comes back for review.
04
Take it with you
Call the same agent from your phone and pick up mid-thought.
05
Give a team a goal
Roles, a budget, and one objective — then they coordinate between themselves.
06
Build on what you know
Earlier research and decisions stay reachable for the next task.
Shared Brain
Good work shouldn’t start over.
Useful research, decisions, and lessons stay close at hand, so the next task starts further along than the last one did.
Meet your Shared BrainPrivacy and control
More capable, still in your hands.
Keep your work close
Run local workflows on hardware you own, and choose which outside services get a look in.
See what is happening
Follow the conversation, read the work, and inspect the result before anything moves on.
Set the limits
Choose permissions and spending limits. Important actions can wait for your approval.
HivemindOS Mini
A little taste,
in the browser.
Small focused tools with nothing to install. Each shows its own pricing and available free usage.
Image Generator
Turn an idea into an image.
Video Generator
Bring a scene to life.
Voice Bank
Find the voice for your next project.
Photo Keyworder
Make your photos easier to find.
Distill
Keep the useful parts of what you read.
Video Transcript
Turn videos into something you can work with.
Where to start
What would you like to get done?
Say it plainly and we’ll point you at a sensible first move.
Start on the hardware you already own
Make this computer your control room.
Install the full app on the computer you use most. Link every other computer as a lightweight worker you manage from there.
Complete HivemindOS Hub
The full dashboard, your lead agent, Shared Brain, approvals, and device controls on the computer you use most.
- Dashboard
- Shared Brain
- Lead agent
- Device permissions
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
- 1
Download and open HivemindOS Link on the extra machine.The guided app installs only the collector, with no second dashboard.
- 2
Click “Link this device.”The app handles installation and shows progress for you.
- 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-onlyPrivate 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?
Hermes, OpenClaw, OpenCode, Codex, Claude Code, Aeon, and MiroShark 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.
Not sure yet?
There are four versions of this page.
Pick what you are trying to supercharge and the demos, steps, and tools above rewrite themselves around it.