HARVEN connects to the security tools you already use, measures your security posture against a built-in knowledge base, and turns the gaps into concrete remediation actions. This page explains what happens under the hood, so you can read your results with confidence and get the most out of every prompt.
The big picture: Measure → Understand → Remediate
Everything in the product is a step in this loop.
The Security Knowledge Base (SKB)
HARVEN ships with a curated Security Knowledge Base: a set of controls (such as Multifactor Authentication, Endpoint Protection, or Vulnerability Management) and, under each control, a set of metrics that quantify how well that control is applied across your environment.
Each metric carries its own success criteria, defined in the SKB. That's why a value turns green or red automatically: the target is part of the knowledge base, not something you have to configure.
Your data sources
HARVEN queries your tools in read-only mode and works with the data where it already lives, rather than copying it into a separate store. When you connect a source, HARVEN reads only the tables it needs to compute your metrics.
Your Context File
Raw connector data rarely tells the whole story: service accounts, guests, legacy systems, naming conventions and business priorities all change how a number should be read. The Context File is where you describe that reality, and it's the single biggest lever on accuracy. The more precisely it reflects your organization, the better HARVEN reconciles raw data with what actually matters to you. (See also Best Prompting Practices.)
The agents
When you add a control, HARVEN doesn't run a fixed script. An orchestrator agent first analyses your workspace : the control you want, the sources you've connected, and your Context File. It then determines which metrics can actually be computed in your environment, and launches a dedicated agent for each one.
Each specialized agent writes and runs a SQL query tailored to your data and your context, then returns the result. This is why generating a control takes a little time : several agents work in parallel. It's also why every metric is fully transparent : you can open any metric to read a plain-language explanation, inspect the exact SQL, browse previous versions, and even edit it.
From insight to action
A metric on its own is a measurement. When one sits below its target, HARVEN generates an action plan and pre-filled tasks you can assign in-platform or push to your existing ticketing tool.
A human always stays in the loop : HARVEN proposes, you decide.
