Space AI is not a service you log into. It is a network you run a piece of — and this page takes you from nothing at all to a machine of your own, watching the places you care about from orbit, while you sleep.
You will meet these on every screen. They are simpler than they sound, and nothing else on this page assumes you know anything.
Nothing is installed in this part, and nothing here is hard to undo. The only step you cannot come back to later is the second one.
Go to spaceai.net/app and press Register. An email address or a wallet is enough — there is nothing to pay and no card to enter.
You are then asked for a password of your own, twice. Choose one you can keep: it encrypts your profile in your browser, before anything reaches a server. That is what makes the data yours, and it is also why we cannot reset it for you.
Leave the cluster field empty. You will land in Open Space Net, the public cluster, which is the right place to start. If you would rather have a private one from day one, type a name for it here instead and you become its admin.
It is shown once. Write it down somewhere that is not the computer you are sitting at.
Because your password never leaves your browser, this code is the only way back to your data if the password is lost. Two minutes now buys you the right never to think about it again.
Everything else on this page can wait until tomorrow. This cannot: a profile with no recovery code and a forgotten password is gone, and there is no support queue that can change that.
Open Detection & Monitoring and add a site — a field, a port, a well pad, a vessel, the roof you are sitting under. The app answers with the next satellite pass over it, and every pass after that.
New accounts arrive with a small ENERGY balance, the credit that pays for imagery. The console shows it as nE. It is enough to watch the thing work; a cluster admin can send you more, and you can buy more when you need it.
You now have an account, a way back into it, and something to watch. Everything below exists for one reason: to keep it all running when your browser is closed.
In a browser, Space AI works while you are there. A node works while you are not — it watches your sites around the clock, keeps your cluster's files replicated, and is the address your hardware reports to. You need one computer that stays switched on: Windows, Mac or Linux, an ordinary desk machine is fine.
The box at the top holds your link and its key, each with a Copy button. This is the same link you will one day send to a colleague — their machine joins exactly the way yours is about to.
Everything after #t= is the cluster key. Anyone holding the whole link can join your cluster, so send it privately and never post it anywhere public.
Paste the whole link into a browser on that computer. The page recognises the operating system and offers the right thing:
From there it runs itself: it asks where to install, sets up Node.js if the machine has none, downloads the node software over a secure connection, makes a desktop icon, starts the node, and opens the app at http://localhost:8777. The cluster and its key are already inside the installer, so there is nothing to type and nothing to configure. Your node finds the others on its own, even behind a home or office router.
The page that just opened looks identical, but it is served by your computer now, not by ours. Sign in with the same account and password. Your profile arrives from the cluster, so your sites, files and balance are already there waiting.
After that the node looks after itself. Every few minutes it checks the network for a newer build, takes it, and restarts. You will never update it by hand.
Close the browser whenever you like — the node keeps working. To stop it entirely, close its window or use the desktop icon again.
None of these are needed to be up and running. Come back to them when you have a reason.
Send a colleague the same link you used, and their machine joins the same cluster. What you mark stays private to you; to share a site, a cluster admin creates a subnet under Manage cluster and you pick it on the site's own panel.
Plug a SPACE AI device in by USB, or power it on the same Wi-Fi. The wizard finds it, updates its firmware, writes the settings onto the device itself, and tells you when its first report lands.
Communications in STORM is where Hubble and Iridium links are switched on for a node that carries the radio. Storage, the 🗄 icon, holds the cluster's files and your own private imagery, replicated across the machines of your cluster; a deleted file stays deleted everywhere.
The icons down the left edge are the apps. They are written in Morse code, one letter each, and every one of them is on both spaceai.net and your own node. None of these steps depend on another; take the one you came for.
Open Spacenger and you see People: everyone with an account in your cluster, who is online, and who has written to you. Choose one, type, press Enter. A red badge on the icon tells you when something is waiting.
Messages are sealed with your cluster's key before they leave your browser, and a thread is only ever between you and one person. A Video call rings the person while their app is open; if they are away, the app says so and you leave a message instead.
Below the people are the nodes and devices. Writing to one of those is different: there is no inbox on the other side, so the thread shows which channel carried the message, the network, LoRa, an SDR or the Iridium satellite link, and whether it went out or is waiting for the node to appear.
STORM is the cloud console. Launch has four kinds of capacity, all bought the same way: pick a size, set where and for how long, read one total in ENERGY and dollars, press the button. You are asked to confirm twice before anything is charged, every time.
A CPU machine arrives with the node installed, hardened with a firewall and automatic security updates, and already joined to your cluster; it is ready to use about five minutes after it starts. GPU prices are read from the market every minute, so what you see is what you rent. Storage and CDN are set up by our team, usually within a business day.
Machines & services is where they live afterwards: you can watch one being built, get root access, stop and start it, add hours, or delete it. An order nothing has been built for yet can be cancelled for a full refund. When the hours you bought run out, the machine is powered off and kept, never deleted, until you add more.
Give it a name and tell it what it is for. Then choose its brain. A private model runs on the agent's own machine, and nothing you say to it leaves that machine. By API uses your own account at a model provider: you paste your key once, it travels to the agent's machine and is forgotten on ours.
Press Deploy, confirm, and watch it come up: building, booting, installing, model loading, ready. A private model takes a few minutes longer the first time, because it downloads. Then talk to it in the panel on the right; it remembers the conversation.
After that you can edit its name, purpose, model or key without rebuilding it, add hours, restart it or delete it. It warns you, in the app and by mail, before its hours run out. Every agent is also a machine in STORM, so root access and the rest are there too.
In STORM, Share & earn is three steps done at your own machine. It shows the node you are at, with its cores, memory and free disk. You switch on what it may share, compute, storage, communications, and move a slider for how much. Compute is shared only while the machine is idle unless you say otherwise.
The last step shows what will be offered and the network's rate for each, in dollars per gigabyte or per hour. Press Publish. Your cluster sees the offer within a minute, and it is written to the ledger. What you share is used only inside your cluster, and you can change it or switch it off here at any time. Files are placed on the machines that stay online: a node is given other members' files while it is available at least half the week (a new node gets a day to prove itself), and a copy on a machine that is mostly off does not count as a copy.
Every node also writes one meter block an hour: how long it was up, what it carried, what it holds for others, and its energy where the hardware can measure it. The Cluster tab shows each node's availability and contribution from those blocks. That record is what you are paid on.
How you are paid. Every day the network settles the previous day: jobs your machine ran, gigabyte-hours it held for other members (while it was online at least half the day and its hold proofs passed), and bytes it served to other members. Served bytes count only as far as the receiving machines wrote a receipt for them in their own meter blocks, so nobody is paid on their own word. You keep 70% of the network's rate. Claim the machine once in your wallet panel (this machine is mine) so its earnings land on your account.
Open your profile and press the ENERGY balance. Every account starts with a trial allowance. To add more, Add funds with USDC: link the wallet you will send from by signing one message, send USDC on Base to the address shown, and the credit appears in your balance within a few minutes, at 1,000 ENERGY per USDC.
The same panel sends ENERGY to any account or node in the cluster, and lists every transaction. Prices across the network are shown in ENERGY and in dollars, so you never have to convert.
Earnings and payouts. If you share compute or storage, the same panel shows what your machines earned, what is owed to you, and the daily settlement roots. Press this machine is mine once on each machine you run, so its earnings are counted for your account. Every Sunday, everyone owed at least 1,000 ENERGY with a linked wallet is paid in USDC on Base; the transaction hash appears next to the payout.
If a colleague sent you a link, you are starting a step ahead — the account already exists and the cluster is already chosen. Read this instead of step 1.
The part after #t= is the cluster key, which makes the whole link a password. Send it privately; never put it somewhere public.
Almost every problem at this stage is one of these, and none of them break anything.
| What you see | What it means | What to do |
|---|---|---|
| “Sign in to use Spacenger”, or a machine order that says “queued” | Your session ended, or the tab is holding an older build than the servers. Orders stay queued only while self-provisioning is off on our side, and they can be cancelled for a full refund. | Reload the page with Ctrl+Shift+R and sign in again. A queued order shows a Cancel and refund button under Machines & services. |
| An agent stuck on “model loading”, or “the agent is not answering yet” | A private model downloads on the agent's machine the first time, up to five gigabytes; after a restart the agent takes about a minute before it listens again. | Wait a few minutes and try again. If it stays there for more than fifteen, restart it from its row; if that does not help, delete it and deploy again, or ask us. |
| “Permissioned cluster” where the join page should be | The link lost its key somewhere between being sent and being opened, or the key is wrong. | Use the complete link, everything after #t= included. Ask whoever sent it for a fresh one if you are unsure. |
| The installer says a download failed | That machine cannot reach spaceai.net. | Check its internet connection and any corporate proxy, then run the installer again. Nothing is left half-installed. |
| “Wrong password” on your own node | Usually a typo. Otherwise the profile on that machine is older than the one you last used elsewhere. | Try once more — the node fetches the newest copy from the cluster as it goes. If it still refuses, use your recovery code. |
| You are in Open Space Net when you expected your own cluster | You signed in on a node that is not in your cluster, or on spaceai.net without a cluster assigned. | Sign in again with the invite link pasted into the cluster field. |
| A device sits at “Configuration” and never moves | The address stored on the device is not one it can reach by itself. | Run the NODE IO wizard again from a computer on the same Wi-Fi as the device. |