Hosting and pricing
Run Community on your infrastructure, choose Team hosting for $649 per workspace per month, or Enterprise from $12,000 per year, plus your volume commitment. Every plan includes collaboration and label history.
Footage: NASA, a heated specimen in a test fixture at the Johnson Space Center materials laboratory, Public domain (NASA) (opens in a new tab), modified: one frame, cropped and graded by Nitsor. No endorsement implied. Sources
Where it runs
Choose who runs your Nitsor instance.
Community is free to self-host. Team includes Nitsor hosting at app.nitsor.com and 250 model-assisted volumes in each monthly billing period. Enterprise adds agreed controls, a yearly volume commitment, deployment assistance and support. Ask for the licence terms and the security documentation before you sign.
The terms at a glance
- Subscription
- Team, $649 per workspace per month, or $6,228 upfront per workspace per year ($519 per month equivalent); Enterprise from $12,000 per year, plus your volume commitment
- Usage
- model-assisted volumes only: 250 per monthly billing period included on Team, then $0.25 each
- Never metered
- reading, reviewing, people, labels, studies, storage
- Read-only people
- included, on every plan
- Caps
- no plan caps on people, projects, labels or versions; fair use applies to storage, action volume and traffic
- Licence
- source-available
- Where it runs
- Community on yours, Team hosted by Nitsor, Enterprise by agreement
In force since 28 August 2026, revised 5, 26, 27, 28 and 30 September 2026.
The document
Our pricing commitments.
These are terms, not features, and they hold on every plan. Each line says the commitment and how you would catch us breaking it. The dated text records them, and your browser computes a fingerprint of the ten headings at the foot.
The licence is source-available.
Nitsor Community is free to self-host under its source-available licence.
You would catch us in the licence terms you sign
Unlimited read-only collaborators, forever.
Anyone who needs to look at the record can look at it. A person who reads and never writes adds no charge, on every plan.
You would catch us on your invoice, next to a collaborator list with no ceiling on it
Reading is never metered.
Opening a scan, reading a label's history and reading a dataset version are included. Charging for them would make the record something you rent.
You would catch us on your invoice, which has no read line
No caps on studies, labels, annotators, projects, branches, Dataset Versions or exports. On any plan.
We do not count or charge for studies, labels, annotators, projects, branches or Dataset Versions. Fair use applies to storage, action volume and traffic; read the fair use policy.
You would catch us on the pricing page, which names no cap
Never per seat, and never per label.
A price per person makes you ration the people who check the work. A price per label makes you ration the work. We do neither.
You would catch us on the pricing page, and then on your invoice
The only meter is scans run through a model on our GPUs.
Community has no hosted compute and no usage meter. Team includes 250 model-assisted volumes in each monthly billing period, one volume being one scan run through one model version, and prices each volume after that at the flat published rate. A run that fails is free. At paid launch, you see and confirm the quote before a run starts.
You would catch us in the Team invoice line items: the upfront subscription and the previous monthly cycle's extra volumes, with their billing dates
Provenance and review state are never behind a paywall.
Who drew a label, who accepted it and what changed are part of the label. Selling access to that would be selling you your own evidence.
You would catch us by opening the history of any label and reading it
Checking a Dataset Version never costs extra.
Reading and checking a dataset version add no charge, on every plan.
You would catch us by checking a version on a machine we do not control
Nothing free moves behind a price later.
Reading, reviewing, people, labels, studies and storage are never metered, on every plan.
You would catch us with this document and its date
Plain export never costs extra and never needs our permission.
Your labels and their history stay readable on every plan, with no approval step in the way. On Team and Enterprise they are readable through the HTTP API. On Community you read them in your own deployment.
You would catch us when you read them, if an approval step or a charge stood in the way
Published by Nitsor · 28 August 2026 · revised 5, 26, 27, 28 and 30 September 2026
SHA-256 of the ten headings above
computing in your browser
Computed in your browser, with no request to us. Save it, and you can tell later whether we changed the wording.
Revision history
- Revision, 30 September 2026
- The bodies of headings four and ten were reworded, and the Caps line above changed. No heading changed, so the fingerprint is the same as on 28 September. Heading four's body no longer says that none of the seven things has a ceiling. It says Nitsor does not count or charge for them, that fair use applies to storage, action volume and traffic, and links to the policy. Heading ten no longer says labels are readable through SDKs on every plan: no SDK is published, and agent credentials are issued in the hosted app, on Team and Enterprise, so it now says the HTTP API on Team and Enterprise and your own deployment on Community.
- Revision, 28 September 2026
- Heading six was reworded for usage pricing. Community is still not metered. Team includes model-assisted volumes each month and prices each one above them on a published scale, and hosted model runs are no longer Enterprise only. The fingerprint no longer matches a copy saved before this date.
- Revision, 27 September 2026
- Headings four, six and eight were reworded for the three plans, Community, Team and Enterprise: "tier" became "plan", Team has a monthly subscription and no usage meter, hosted model runs are part of Enterprise, and a paid plan includes rather than frees. The fingerprint no longer matches a copy saved before this date.
- Revision, 26 September 2026
- Headings eight and nine were reworded for the one hosted plan: they had named an approval step and a Core tier. The fingerprint no longer matches a copy saved before this date.
- Correction, 5 September 2026
- Two headings changed, so the fingerprint no longer matched a rev 2026.08 copy. Line eight had said signing a release stayed in Core, and line ten had said export was complete and unconditional. Both were narrowed: checking stayed free, and the export line bound the price rather than the feature set.
The meter
One meter: scans run through a hosted model.
Model runs are priced per completed scan/model result, never per person. Team hosting costs $649 per workspace per month, or $6,228 upfront per workspace per year ($519 per month equivalent), and includes 250 model-assisted volumes. Each volume after that costs $0.25 in each monthly billing period on both Team plans. Adding people or labels adds no fee.
These are launch prices. Paid checkout and usage charging are not active. The subscription is paid upfront. Extra usage is settled after each monthly billing cycle. A cycle follows your billing dates and can start partway through a calendar month. A workload outside the supported models and scan sizes, including an exceptionally expensive run, needs a separate fixed quote agreed before it runs, with no surprise runtime charge.
Paid upfront one thing
- The subscription, $649 per workspace per month, or $6,228 upfront per workspace per year ($519 per month equivalent)
Settled after each monthly cycle one thing
- Model-assisted volumes above 250 in a monthly billing period
Never metered eleven things
- Reading anything
- People who only read
- Seats
- Labels
- Studies and series
- Annotators
- Projects and branches
- Dataset versions
- Reviews
- Provenance and review history
- Your own storage provider, when you connect customer storage
Show us what you inspect.
Start with a public CT or MRI scan. See how a model suggestion becomes a reviewed label and a fixed training dataset.