On this page
Short answer: self-hosting suits teams that must keep project data on their own infrastructure, or that want control over upgrades and network access, and that have someone able to run a server. It does not suit teams with nobody to patch, back up and monitor it. The software price is only part of the cost. Plan for the time too.
What "self-hosted" means
You run the application on a server you control: your own data centre, a virtual machine in your cloud account, or a PC on your network. You own the database, the backups, the updates and the uptime. The vendor does not see your data unless you send it.
It is different from a vendor-hosted tool in a "private cloud" or "dedicated instance". There the vendor still operates it. Ask which one you are being sold.
Benefits
- Data location. Project data, which often names clients, budgets and risks, stays on infrastructure you choose. That helps with residency rules and internal policy.
- Network control. It can run behind a firewall or on an air-gapped network, and talk to internal systems such as Jira, Bitbucket or CI servers that are not reachable from the internet.
- Upgrade control. You choose when to upgrade, test first, and avoid surprise changes.
- Cost shape. Costs follow your server and licence, not a per-seat cloud plan that grows with every new person.
- Identity on your terms. Single sign-on, MFA and audit logs tied to your own directory.
Drawbacks, honestly
- You run it. Backups, restores, patches, certificates and monitoring are your job. A backup you have never restored is a hope, not a plan.
- Slower start. There is an install, configuration and often a network or security review before the first user logs in.
- Updates are manual. Fixes and features arrive when you apply them.
- Support is narrower. If the server is down at 3 a.m., it is your problem.
- Scale and availability are yours to design. A single server is a single point of failure unless you plan otherwise.
Compare the real cost over three years
Do not compare a licence price to a cloud price. Build a simple model with your own numbers:
| Cost | Self-hosted | Cloud |
|---|---|---|
| Software | Licence or support fee | Per-seat subscription |
| Infrastructure | Server or VM, storage, backups | Included |
| Admin time | Hours per month, times your hourly cost | Close to zero |
| Setup | Install and security review, one off | Close to zero |
| Growth | Usually flat, then a licence tier | Linear with users |
Write down your assumptions (number of users now and in three years, admin hours a month) and compute the three-year total for each. The result often flips depending on team size: small teams usually favour cloud on cost, and large or regulated ones often favour self-hosted.
Questions to ask any self-hosted vendor
- Which databases are supported, and is the default one fit for production?
- How do backup and restore work, and has the restore been tested?
- How does an upgrade work, and can it be rolled back?
- Does it work offline or on an air-gapped network, including licence activation?
- What leaves my network? Ask about telemetry, licence checks and any AI features.
- Is there single sign-on, MFA and an audit log with export?
- What happens to my data and my access if the vendor disappears?
How NUSK is deployed
NUSK runs three ways: Docker Compose (backend and frontend containers), a desktop installer for Windows, macOS and Linux, and a compiled server image for the on-premise edition.
- Database. SQLite by default. PostgreSQL, MySQL, Microsoft SQL Server and Oracle are supported for production.
- Backup. A script snapshots the database. An admin page downloads a point-in-time backup while the app is live, or a zip with the database and every uploaded document. Restore means stopping the app and putting the files back.
- Network. It can run on a LAN, and reach internal Jira, Confluence, Bitbucket and CI servers.
- Offline and air-gapped. The on-premise edition supports an offline signed licence file for sites with no internet access.
- What leaves your network. A licensed server checks in with the vendor and sends only an installation ID, a machine fingerprint, the host name, the version and the active user count. It never sends project data. With AI turned off, nothing is sent to an AI provider.
- Security. Single sign-on with Google, Microsoft or any SAML 2.0 provider, SCIM provisioning, two-step verification, an audit log with CSV export and SIEM streaming.
Before production, set a long random JWT_SECRET and use HTTPS. The installation guide covers both.
When NUSK is not the right choice
- You have no one to run a server and want a service that just works. Choose a hosted tool, or NUSK's hosted edition.
- You need a large marketplace of third-party add-ons.
- You need a certified compliance report from the vendor. Check what we hold before you assume.
Try it
Run the Docker version on a spare VM, import one real project, and restore a backup on a second machine. If that goes smoothly, you will know what self-hosting really costs you.
Request the on-premise edition · Start free on the hosted edition
See it with your own project
Import a plan or a Jira project into a free NUSK workspace and judge it on your own data.