Blog

Product

Introducing DefenseOS: The Operating System That Runs Your Cyber Defense Factory

AI agents are only as effective as the context and runtime they operate in. DefenseOS gives Cogent's agents both, so your vulnerability and exposure management program runs at machine speed under rules your team sets.

5 minutes read

Josh Dreyfuss

Marketing

Josh Dreyfuss

Marketing

The threat landscape has fundamentally changed. In October, Microsoft fixed a record 1,416 security weaknesses in a single Patch Tuesday, more than in all of 2025 combined. Cogent's Q4 2026 Threat Report found an average of 34 new attack paths per month that AI agent swarms could exploit, up 386% year over year. The volume of vulnerabilities has exploded while the time to exploit has shrunk to hours.

Security teams need to move at machine speed to keep up, and that means finding ways to leverage the power of AI agents for defense. But an agent without context barely helps: it treats every critical CVE the same, because it doesn't know which assets matter, who owns them, or how your team works.

Today we're spotlighting the DefenseOS, the context engine and AI harness at the core of the Cogent platform. We’re also introducing two new capabilities that play a key role in DefenseOS: Agent Policies, which codifies how your organization works, and Agent Automations, which runs recurring work on its own. Together, they help lay the foundation for the Cyber Defense Factory.

Machine speed takes more than a smarter model

Putting AI agents to work across the full vulnerability lifecycle runs into some major issues that cannot be overcome without a platform purpose-built to support them. 

The first is a lack of  shared context. AI agents must have access to a unified, living set of context that represents an organization’s assets, findings, exposure, ownership, and ways of working. That knowledge lives in CMDBs, security scanners, policy documents, and people's heads, so most security teams only see a fraction of it. Manually bringing this together in spreadsheets or in the form of handwritten static rules is too slow and incomplete for agents to work effectively. A smarter model discovers rules itself and surfaces them to security teams to review, and continuously checks against drift and stale information. 

The second is an ill-defined and scoped runtime environment. a purpose-built AI harness. AI agents need a runtime environment that is designed from the ground up to empower them to do effective cyber defense work across investigation, prioritization, exposure validation, verified remediation, and reporting. This means guardrails and guidance based on  limits the security team sets, clear and transparent evidence behind every decision, and a handoff to people when confidence is low. Without that, few security leaders will let agents do more than suggest.

The third is a lack of automation and ability for security teams to direct agentic work. Analysts pull data from half a dozen tools into spreadsheets for the weekly report, then rerun the same investigations the next week. AI assistants help, but someone still has to remember to type the prompt. If security teams cannot easily direct AI agents to automate repetitive work and run clear workflows, then there will always be limits on how effective they can be. 

Introducing DefenseOS

DefenseOS is the operating system for AI agents doing defensive work in your environment. It does two jobs.

A context layer that learns your environment

DefenseOS ingests data from scanners, business systems, and security tools through more than 100 integrations, with no scanner replacement required. It builds a deep understanding of your environment and the ways you work, normalizes and deduplicates findings, then links every vulnerability to the assets, owners, and dependencies it touches. Agent Policies sits on top, turning what DefenseOS learns into rules you can read, edit, and approve.

An execution layer built for governed autonomy

DefenseOS is also the AI harness our agents run inside. Approval gates, confidence thresholds, and policy limits decide what an agent may do on its own. Every action is logged and traces back to its recommendation, and low-confidence calls escalate to a person. Most teams start with agents recommending fixes that people approve, then widen autonomy as results hold up. Agent Automations extends this layer.

Figure 1: DefenseOS sits beneath every stage of the Cyber Defense Factory.

Agent Policies: how your organization works, discovered and codified autonomously

Every enterprise runs its security program differently. To make vulnerability management effective, you have to find a way to capture and codify that. For years, that’s been done manually by connecting data sources with spreadsheets and CVSS scores. Then a wave of tooling came in that ingested all those data sources and let you build rules to dictate context by hand. This was an improvement, but still manual and brittle, as you had to create the rules directly and maintain them yourself. 

Agent Policies is the next wave of codifying context. Agent Policies autonomously reasons through and understands how your organization functions and what matters to you, and surfaces those as rules that shape how Cogent handles asset context, risk scoring, prioritization, and remediation. A policy might mark the assets behind your payments platform as tier 1, or establish how to route findings on a network segment to the team that owns it. Agent Policies builds and maintains rules in three ways:

  • Rule discovery. Agents propose policies from what DefenseOS finds across your environment. You accept, edit, or reject each one before it takes effect.

  • Rule creation. Add policies directly by uploading documents or describing context in natural language.

  • Drift detection. Agents spot drift and new context and propose updates, so policies stay current.

This approach not only saves significant time and effort by making context capture autonomous and faster to implement, but it helps organizations get a much more accurate picture of their environment, as the agents will propose new rules that your team may not have considered or that would have gone overlooked in a dynamic environment. 

Agent Automations: define recurring work once, let agents run it

Much of a security team's week goes to work it has done before: the weekly exposure report, a recurring investigation, the alert when a critical CVE hits a key system. Agent Automations turns that work into agent workflows that run on their own.

  • Control when automations run. Set a schedule or trigger on events in your environment.

  • Start fast. Pick from a library of 30+ templates or build your own in natural language.

  • Scale across the organization. Share proven workflows across teams and business units.

  • Send results where your team works. Connect automations to Slack or email to deliver reports and notify teams when findings need attention.

By automating known, repeat workflows, you can free up time to spend on more critical vulnerability and exposure management efforts. And because automations run on DefenseOS, triggers can fire on events only Cogent can see, such as a new finding on an asset your policies mark as business-critical. One enterprise customer set up nine reporting automations, replacing hours their team used to spend on assembling those by hand.

From Zero Day to the Friday report

One large financial institution we work with ranks teams by their distance from the CEO and treats assets closer to the top as more critical. That rule never lived in a scanner. Agent Policies captured it, and the security team approved it. Now picture the next major zero day. 

  1. New zero day is disclosed. Zero Day Response in Cogent detects the new zero day in minutes. DefenseOS matches it to affected assets and owners using scanner, CMDB, and network data.

  2. Policies set the priority. Asset criticality rules like the CEO-distance rule and internet exposure push the 5% of affected assets that are most critical to the top of the queue. 

  3. An automation fires. This new zero day triggers an agent to investigate and alert the owning team through Slack or email with remediation guidance.

  4. The fix is verified. The fix runs at the autonomy level the team approved, and a rescan confirms it's closed.

  5. The report writes itself. On Friday, a scheduled automation rolls up what changed for leadership.

The team set one policy and two automations. DefenseOS and Cogent's agents carried the rest.

Your own Cyber Defense Factory

The Cyber Defense Factory runs security like a production line: inventory, discover, validate, remediate, verify, with each verified outcome feeding the next run. DefenseOS is the floor that line runs on. Fortune 500 security teams running Cogent have cut the exposure window for critical vulnerabilities by 97%, and customers have tripled the output of their vulnerability management teams.

See DefenseOS in action: Book a demo to see how it works, and dive into Agent Policies and Agent Automations.

Related articles

View all articles

View all articles

B8oAo#kI  aZ  dBeRmGoF

See Cogent In Action

Schedule a personalized demo today to learn how Cogent can supercharge your vulnerability management program.

Book a demo

Book a demo

Free risk assessment

Free risk assessment

BToPoEk@  aZ  d@e#mWo2

See Cogent In Action

Schedule a personalized demo today to learn how Cogent can supercharge your vulnerability management program.

Book a demo

Book a demo

Free risk assessment

Free risk assessment

BDo4oTkN  aB  dJe3mUoK

See Cogent In Action

Schedule a personalized demo today to learn how Cogent can supercharge your vulnerability management program.

Book a demo

Book a demo

Free risk assessment

Free risk assessment