Pricing
You pay for the machine. Tokens are never billed.
Introductory pricing, honestly arrived at: this round runs on 2024-generation servers, bought before memory prices went vertical, your price is locked for 12 months, and the next round's prices are posted below — so you can see exactly what you're early to.
Why these prices exist, and when they end
We buy servers in rounds, and we price each round off what its hardware actually costs us — the memory above all, but also the network the machines ride. The round open right now runs on 2024-generation servers, bought before memory prices went vertical — which is why the prices below are low, and why we can say so plainly instead of calling it a promotion.
Join while this round is open and your price is locked for 12 months. No surprise emails, no mid-year increase — we only ever change prices when our hardware costs change, because that is the only change we could honestly justify to you.
When we can no longer buy this hardware at this price, this round closes to new signups and the next opens at the prices posted on every card below. Memory prices are climbing across the whole industry; ours will always be predictable, because they are never ours to invent.
Accounts are free. Servers are not — but the first week is.
A free account is exactly that — sign in, look around, configure your Host, import an agent. Nothing is running yet, and we won't pretend otherwise. Your server exists when you pick a plan, and from that moment it's on around the clock, whether or not you are.
Your first plan starts with a 7-day free trial — a running server, on the plan you chose, nothing charged until the week is up. We ask for a card up front because a free machine with an internet connection is what gets abused; cancel inside the week and you are never charged.
United States customers only, for now. We'll say so here when that changes.
Personal plans — this round's pricing
- Starter$10/monext round: $13/mo7-day free trial on every plan
Your always-on host, sized for one person and a couple of apps.
1 GB resident memory · 5 GB storage- 1 GB memory — your host and a couple of open apps
- 5 GB storage
- 10 sandbox vCPU-hours and 100 GB egress a month, included
- Uncapped CPU burst into spare capacity
- Publish endpoints to the network
- Solo$19/monext round: $25/mo
Your always-on host. Everyday use, one flat price.
2 GB resident memory · 10 GB storage- 2 GB memory — your host and a handful of open apps
- 10 GB storage
- 25 sandbox vCPU-hours and 250 GB egress a month, included
- Uncapped CPU burst into spare capacity
- Publish endpoints to the network
- Pro$34/monext round: $44/mo
More room, priority, your own domain and private network.
4 GB resident memory · 25 GB storage- 4 GB memory — more apps and sidecars open at once
- 25 GB storage
- 60 sandbox vCPU-hours and 500 GB egress a month, included
- Priority when capacity is contended
- Custom domain
- Private network — bastions and relays for your own machines
- Pro-Heavy$72/monext round: $92/mo
The most room, the highest priority.
8 GB resident memory · 50 GB storage- 8 GB memory — a full workspace open at once
- 50 GB storage
- 120 sandbox vCPU-hours and 1 TB egress a month, included
- Highest scheduling priority
Every price is locked for 12 months from the day you subscribe. Pay for the year instead and it's 10% off — $205/yr for Solo, $367/yr for Pro. Switch plans any time, pro-rated against the day. Cancel and the unused time goes back to your card automatically — two weeks in, two weeks back. Refund policy. Memory is the number to watch: it's how much your server holds open at once. CPU is uncapped burst on every plan. Pro is deliberately the best value per GB on this ladder — it's the household plan, two real hosts.
Above 8 GB, memory pools
The personal ladder ends at 8 GB on purpose. Above it you're running hosts for a team, for clients, for a business — so capacity stops being a plan and becomes a pool: one block of memory across all your hosts, allocated by you, balanced automatically, hosts themselves free and unlimited. Three ways to hold it:
- Pooled$12/GB + $20/seatnext round: $15/GB
From 16 GB, on our network. Add hosts free, size them from the pool, scale any of them to zero. Seats are the people on the account.
- Dedicated$15/GB reservednext round: $19/GB
Your pool on reserved hardware — nothing shared, nothing borrowed, run by us. A reserved GB costs more than a pooled one because it's held for you around the clock. That's the honest price of isolation, not a volume discount that never comes.
- License$2.50/GB + $15/seatnever changes with hardware rounds
The platform on hardware you bring — a server you rent anywhere, or metal you own. From 16 GB. There is no hardware of ours underneath this price, so it never moves and never closes.
All three onboard through us for now: info@privatae.ai.
What each personal plan includes every month
| Plan | Memory | Storage | Sandbox compute | Egress | Can run |
|---|---|---|---|---|---|
| Starter $10/mo | 1 GB | 5 GB | 10 vCPU-hours | 100 GB | — |
| Solo $19/mo | 2 GB | 10 GB | 25 vCPU-hours | 250 GB | Browser |
| Pro $34/mo | 4 GB | 25 GB | 60 vCPU-hours | 500 GB | Browser |
| Pro-Heavy $72/mo | 8 GB | 50 GB | 120 vCPU-hours | 1 TB | Browser |
Memory is what your host can hold open at once — the host itself, its database, its private network, and any app you have running. Some apps are big: the Browser holds a real browser open while it works, so it needs a GB to itself. That is what the last column means. You can install any app on any plan and set it up with your host; a job that needs more memory than your plan has spare will not run until you move up. Nothing is hidden and nothing runs up a bill.
Past the included amounts: sandbox compute at $0.05 per vCPU-hour, per second; egress at $0.05 per GB. Both are added to your monthly bill, at the same rate on every plan. Storage is the disk your server is built with — need more, move up a size. Your own server is never metered.
Every plan includes
- Always-on host — your own machine, never metered
- Sovereign identity — your host holds its own keys and certificate
- Personality and persistent memory
- Any model, any provider — your keys, we never bill tokens
- Your own private network — publish and call endpoints over mTLS
- Sandbox compute and egress included every month; one published rate past that
- Automatic container updates
- 7-day free trial, then cancel any time — unused time refunds automatically
One metered rate
Some jobs want more machines than yours. Build sandboxes, editor sessions, dev servers — your server pulls them on demand, they run as separate machines, and every plan includes a set number of hours of them each month. Past that they meter per second at $0.05 per vCPU-hour and go on your monthly bill. Stop them and the meter stops. Every plan pays the same rate and can pull as many as it wants: the plan sizes your machine, not your ambition.
Your own server never meters. There is nothing to top up and no balance to run dry — usage past what your plan includes is billed at renewal, and nothing else ever is.
Public sites
Publishing to the public side is built into every paid plan, and dev previews are included. A production site is $20 a month, which buys the platform around it: hosting, a managed database, authentication, and billing — with your customers in your tables and your revenue in your own Stripe account. We take no cut of what a site earns.
That $20 carries reasonable usage limits, because bandwidth costs money and "unlimited" is how platforms lie to you. The limits fit a working business; outgrow them and we'll talk.
What you never pay extra for
- Tokens. Inference and embeddings run on your own keys and endpoints. There is no token meter to mark up.
- Bursts. CPU spikes ride into spare capacity. Your own machine is never billed by the minute.
- Updates. The container stays current automatically.
- Your own data. Memory recall, the audit log, backup status — reading what's yours is not a feature tier.
Common questions
- Why is this called introductory pricing?
- Because the discount is economics, not marketing. This round runs on 2024-generation servers, bought before memory prices went vertical. While we can still buy that hardware at that price, these prices hold. When the market takes it away, this round closes to new signups and the next opens at the posted prices. Existing members are unaffected.
- How long is my price locked?
- 12 months from signup. Inside the lock your price cannot change — no surprise emails. We only ever change prices when our hardware costs change, because that is the only change we could honestly justify to you.
- What does the monthly price cover?
- Your always-on server: its memory, storage, network edge, and the certificate and audit infrastructure that keeps it yours — plus a set amount of sandbox compute and egress every month.
- Is there a free trial?
- Yes — 7 days on your first plan, card required, nothing charged until the trial ends. Cancel before then and you are never charged.
- Who can sign up?
- Customers in the United States, for now. We'll say so on this page when that changes.
- How do I pay for the model?
- You already do — through whatever provider you already use, your own key, or a model you host yourself. Any model, any source. We are never in that billing path. Where a subscription applies, it is used through the provider's own tools and standard sign-in, under their terms of service — we never build a login of our own.
- What happens past the included usage?
- Sandbox compute past your plan's included hours meters at $0.05 per vCPU-hour, per second, and is added to your monthly bill. Egress past the included amount is $0.05 per GB. Nothing to top up, nothing stops mid-job. Your own server is never metered.
- What if I need more than 8 GB?
- Above the personal plans, memory is pooled — see the three structures above: pooled on our network, dedicated on reserved hardware, or licensed on hardware you bring. The license price has no hardware of ours underneath it, so it never moves with hardware rounds and never closes.
- What does a public site cost?
- $20 a month per live site — hosting, managed database, auth, and billing included, with sensible usage limits. Dev previews are free on every paid plan.
Bigger questions? info@privatae.ai