Monzy MerzaCo-founder & CEO

Our mission is to make every security practitioner as effective as the entire team. This standard enables us to focus on solving the hardest problems in security operations.

Crogl acts as an AI SOC agent that enables security teams to hunt threats and investigate alerts using all their existing data without having to move it.

Historically, vendors told customers to put all data in one place, but that is not the reality. Data lives in various data lakes, databases, and cloud providers, leading to fragmented, non-schematized information.

Crogl allows analysts to conduct deep security investigations without forcing them to normalize or schematize data across all different tools or stores.

We went from absolute distrust, to teams attempting to build their own systems, and finally to realizing that building enterprise-grade AI SOC agents requires a highly engineered solution.

A chatbot does not provide consistency, governance, auditability, or sustainability. True SOC operations require routing between multiple models, context, and consistency.

Alert fatigue is driven by tool and data fragmentation, expecting practitioners to hold a wide breath of domain and database competencies, and a growing enterprise footprint with evolving threats.

Security is the only industry where we expect one practitioner to be an expert in database queries, threat analysis, IT stacks, and regulatory compliance. It is an untenable width of expectations.

It is a three-part approach combining a reasoning system (the LLM), a rules-driven deterministic system (a knowledge graph for auditable consistency), and a harness (for integration, safety, and governance).

We serve critical environments like utility companies, banks, and government agencies. These environments have stringent security controls and cannot use typical cloud-based SaaS solutions.

We suggest starting interactively to build confidence, then deploying Crogl autonomously for a single use case, and gradually opening the aperture from one use case to another.

The volume of alerts is too high for humans to inspect every single query. Teams must let the AI work and instead focus on building systems to govern that automated work.

Humans can decide on query execution, ticket-level escalations, or act as the specification writer defining the AI's playbooks and success criteria. The third degree is the most superior.

Given the mixed messages in the industry around security AI, we decided to put the product in a private preview for the community to test and decide for themselves.

It allows users to share transparent, open feedback. Our engineering team, product teams, and I are directly active in the Slack channel to help users troubleshoot and build.

It is built for highly consequential environments—like nuclear plants, manufacturing sites, utility grids, and major banks—that need secure, robust, and air-gapped capabilities.

We will see widespread adoption of AI technologies, a rise in AI-driven attacks like the Mexican government breach, and sophisticated cyber campaigns run for under eighty dollars in tokens.

Get your hands dirty and try it. Just like the security community did with Snort or Yara, download the tool, tinker with the capturing, and participate in the project.