Organisation policies (coming)
Coming. Organisation policies are designed but not built yet. This page says what to expect.
Out of the box, Kaviris is permissive about your own choices: a vulnerable library is a recommendation, never a blocker. Some organisations want stricter rules for all their apps, for example to meet NIS2. Policies let them opt in.
What a policy can say
- Which outside services apps may reach (say, only hosts in the EU).
- Which ports and kinds of traffic are allowed (say, no plain HTTP).
- Which vulnerabilities block a release (say, anything known to be exploited, or anything rated 7 or higher that has a fix).
- Which licences and packages are allowed.
Each rule is set to one of:
| Mode | Effect |
|---|---|
| enforce | Blocks a deploy or contract change that breaks it |
| warn | Never blocks; shows on the deploy, the app card and to your agent |
| audit | Only recorded and counted: a way to try a rule before enforcing it |
Groups
Apps can be put in groups inside the organisation, each with its own extra rules. A group can only make things stricter, never looser than the organisation.
Ready-made packs
Kaviris will offer packs you can add instead of writing rules yourself: for example "block known-exploited vulnerabilities", "sensible defaults", "EU-only outbound traffic", a NIS2 starting point and licence lists. A pack helps; it is not a compliance certificate.
What your agent sees
A deploy blocked by a policy gets a refusal like any other: which rule, why, and how to fix it. The agent can read the effective policy and explain a refusal. If a rule can't reasonably be met, it can ask for an exception, but only with your agreement.
An exception is narrow (one host, one package version, one vulnerability), time-limited (it must have an end date), and approved by an owner or admin in the portal, who can't be the person asking. Every step is recorded.
Who sets policies
Owners and admins, in the portal: write and change rules, try them out first (a dry run shows which apps would be blocked), keep a history of versions, and approve exceptions. No coding agent can change a policy.
Nothing a policy says can switch off the platform's own protections: outbound traffic denied by default, one micro-VM per app, only signed images, and refusing secrets in code.