Deployment models

Three deployment models.
One product.

You decide where scanning happens and where the results are kept. All three models run the same runner with the same engines; the only difference is who operates it.

  • Model 1Online

    Dashboard and runners run at Kodator. You connect your Git hosting and start.

  • Model 2Hybrid

    The dashboard runs at Kodator, the runner in your network. Your code stays in-house.

  • Model 3On-premises

    Dashboard, database and runners run at your site. Findings and evidence stay in-house too.

Comparison

The models compared

Comparison of the deployment models online, hybrid and on-premises
CriterionOnlineHybridOn-premises
Dashboard and APIHosted by KodatorHosted by KodatorAt your site, on your own server or your own Kubernetes
Runner with the scan enginesHosted by KodatorIn your network, the same runner imageAt your site
Where is your source code?The runner at Kodator fetches it from your Git hosting into a short-lived working directory for the scan. The findings are stored, not the repository.It does not leave your network. The runner scans on site.It does not leave your network.
Where are findings and evidence?At Kodator: findings, SBOMs and audit log in a separate database per customer.At Kodator, in a separate database per customer. The runner sends only findings, SBOM and metadata.At your site, in your own PostgreSQL database and your own storage.
Who operates and updates?Kodator: dashboard, runners and engines.Kodator the dashboard. You the runner: a new image version brings the new engine versions.You. Updates arrive as new image versions; the database is migrated automatically during the update.
NetworkKodator needs access to your Git hosting (app or access token). Web applications have to be reachable from the internet for runtime testing.Outbound HTTPS from the runner to Kodator only, no inbound ports, proxy possible. The runner downloads vulnerability databases from the internet.Outbound access to public vulnerability databases and the databases of the engines.
InstallationNoneOne container, started with Docker, Docker Compose or the Helm chart for the runner. Registration with a one-time token from the dashboard.Docker Compose (PostgreSQL, API, runner, dashboard) or the Helm chart for Kubernetes.
Internal Git serversOnly if they are reachable from the internet.Not connected directly yet. Today: scan in your own pipeline and upload only the results.Yes: GitHub Enterprise Server, self-managed GitLab, Azure DevOps Server.
Testing internal web applicationsNoYes, through the runner in your network.Yes
SuitsTeams that want to start quickly and host their code in the cloud anyway.Banks, regulated companies and manufacturers whose code must not leave the premises.Organisations that want or have to keep findings and evidence themselves as well.

In detail

What you should know

  • One runner image for all models

    The runner at Kodator and the runner in your network are the same image. The engines in it are pinned to fixed versions and verified by SHA-256. An engine update is always a new image version: traceable and without surprises.

  • What leaves your network in the hybrid model

    Findings, the SBOM and metadata about the scan. Credentials and card numbers in findings are masked before storage. To be open about it: one of the engines queries a public vulnerability service with package names and versions. A switch to turn this off is planned.

  • One database per customer

    In hosted operation every customer gets a PostgreSQL database of their own. A query cannot return data of another customer because it is not in the same database. When the relationship ends, exactly this one database can be deleted.

  • Scanning in your own pipeline

    If you do not want to give Kodator access to the repository, you let the runner image scan inside your own CI/CD pipeline and upload only findings and the SBOM. A connection to the repository is not needed then.

  • Networks without internet access

    Vulnerability data can be carried into an isolated network as a bundle with a SHA-256 checksum. Fully offline operation of all engines, including their own databases, is planned and not available today.

  • Updates when you operate it yourself

    On-premises you install new versions yourself, with Docker Compose or Helm. The databases are migrated when the new version starts. Kodator updates the vulnerability data itself once a day, provided the server can reach the sources.

Where things stand

Kodator is a young product. The hosted environment in the EU is being set up; until then we show Kodator in a demo environment. Talk to us about which model fits your start.

Next step

See Kodator on your own code.

Tell us which product and which requirement you are dealing with. In a demo we show you which evidence Kodator supplies for it, and which it does not.