Kelvar Docs

Get started

Kelvar documentation

Kelvar is a set of always-on security agents. They read your code and your alerts, record only what they can confirm, and hold a merge when something critical is found. This documentation explains how it works, how to set it up, and what it costs.

#What Kelvar does

Kelvar works in the gap between a scanner and an analyst. A scanner finds patterns. An analyst decides whether they matter. Kelvar agents do both, and they show their evidence.

  • They audit. An agent reads source files, dependency manifests, workflows and alerts, inside a sandbox that is limited to the paths you admit.
  • They record. Every confirmed finding carries a file, a line, an advisory and a remediation, and it lands on the Findings screen.
  • They intercept. A critical finding can hold a merge before it reaches production, instead of sending an alert afterwards.
  • They prove. Every action is a row in an append-only ledger. A fact is only marked true after a deterministic check passes, never because a model said so.

#Two ways to use it

ScannerAgents
What it isA fast, deterministic command-line checkModel-driven audits in the Kelvar app
Best forCI gates, pre-commit checks, quick triageDeep audits, triage, patches, reports
Needs a modelNoYes, your keys or ours
Billed asScan runAgent run or Deep run

Most teams start with the scanner in CI and add agents for the work a scanner cannot do.

#Where to go next

New to Kelvar? Start with the Quickstart, then read How Kelvar works. If you are evaluating Kelvar for a security review, read the Security model and Permissions and sandboxing.