AI Agents Hub | AI triage across CloudSEK products

AI agents investigate first.
Your analysts make the call.

Agents Hub investigates findings from CloudSEK products, gathers the evidence, and hands your analysts, and the tools in your SOC, a recommended verdict with the reasoning attached.

Agents run checks in parallel
One evidence record per finding
No auto-remediation or auto-closure
Evidence your SOC tools and AI assistants can act on
1 · Findings arrive from every product
Employee credentials in a stealer logXVigil · Credential breach
Node.js app on a build open to Node2ShellBeVigil · Web App Scanner · CVE-2026-111497
MCP server reachable from the internetAIVigil · AI event
Cloud API key in a public repositoryXVigil · Code repositories
Linux host on a kernel open to CVE-2026-98432BeVigil · Network Scanner
Public API collection names your APIBeVigil · API Scanner
Dark web, code, credentials, domains, attack surface, CVEs, AI infrastructure, APIs. More sources as products are added.
2 · Agents Hub checks each one
Agents Hub

Specialised agents check each finding before it reaches your team.

Checking: stealer-log credentials
  • Is it real, and still live?Safe, non-destructive checks
  • Is it yours, and what can it reach?Asset, owner and context
  • How much does it matter?Severity and likely impact
  • What should happen next?Verdict, reasoning and next step

Noise is set aside, not deleted. Real findings move on with their evidence.

3 · Your SOC gets the insights and decides the actions
Your SOC

Each finding arrives with a verdict, what it exposes and a recommended action, ready for your analysts and the tools they run.

  • Actionable3 employee logins still workReset the passwords and enforce MFA
  • ActionableUnpatched, listed in CISA KEVUpdate Node.js, block Brotli bodies meanwhile
  • ActionableNo auth, tools open to anyoneRestrict access and require auth
  • ActionableKey valid, opens cloud storageRotate the key and audit access
  • PotentialFix pending, no public exploit yetSchedule the kernel update
  • Non-actionableReturns no sensitive dataSet aside, kept on record
Nexus AICorrelates related findings across products, quantifies the risk, and answers questions in plain language.
Analyst time and AI tokens go to real threats, not first-pass checks.
10–30 min

of manual triage per finding, handled by agents

3 verdicts

Actionable, Potential or Non-actionable, each with its evidence

4-step record

observation, validation, context and recommendation for each triaged finding

Why it matters

The first pass on each finding is still manual.

Before anyone can act on an external finding, someone has to run the same first checks. Today that someone is an analyst.

01 / VOLUME

Repetitive first checks

Open the page, follow the redirect, read the code, test whether it is still live. Then do it again for the next finding.

02 / CONSISTENCY

Different calls on the same finding

Two analysts can reach different conclusions on the same finding, and the reasoning is rarely written down.

03 / PRIORITY

Severity buried in the queue

High-severity findings wait in the same queue as noise until someone reaches them.

04 / TRUST

Automation that cannot show its work

When a tool cannot explain its verdict, analysts redo the work or set the result aside.

05 / REUSE

Every tool redoes the same investigation

Your SOAR, your SIEM and your AI assistant each start from the raw finding, so the same checks run again, with the same tokens and the same wait, before anything can act.

Finding Module Status
Findings sit in arrival order. Nothing tells an analyst which one to open first.
How it works

Agents run their checks side by side. The Decision Agent turns the results into a recommendation.

Each agent family covers one kind of finding. The Decision Agent correlates what they return, so analysts do not repeat the checks.

One finding, investigated in parallelIllustrative flow
Select an agent family for detail
InputFindings
XVigilBrand, code, credential and API findings
BeVigilWeb, network, API and CVE findings
AIVigilAI events: exposed AI infrastructure
More products as support expands
Specialised agentsRun side by side
Decision AgentCorrelates and recommends
Evidence received
  • Web and visual evidence
  • Credential and form evidence
  • Code and API evidence
  • Attack surface evidence
VerdictActionable
ConfidenceHigh
Next actionApprove takedown
Selected family

Web and visual agents

Browse suspected phishing and lookalike pages, follow redirects, capture screenshots and compare them against your live brand site.

What it checks

URLsRedirectsScreenshotsBrand match

How it fits

Runs on its ownChecks start without waiting for other agents.
Feeds one recordEach output goes into the cross-module investigation record.
Hands off to an analystA family returns evidence, not actions.
One finding, up close

Follow a finding from first signal to recommendation.

Pick a use case or let it run. Each investigation records what was found, what was checked, why it matters and what to do next.

Example investigationIllustrative run
Step 1 of 4 · Observation
Lookalike domain observed
Investigating

A domain resembling the monitored brand resolves to a live page. The signal and its location are recorded.

Evidence
Evidence and recommendation go to your analysts
Video walkthrough

Watch an investigation move from signal to recommendation.

See agents gather evidence, correlate it across modules and pass a complete record to the analyst.
Coverage

One investigation standard across CloudSEK products.

Coverage grows as products and modules are added. Every module records the same four steps for each finding, so verdicts read the same way across XVigil, BeVigil and AIVigil.

Fake URLs and phishing

Captures screenshots and redirect chains, and compares the page with your live brand site.

Xvigil

Code repositories

Separates your internal code from public noise and checks, in a sandbox, whether an exposed key still works.

Xvigil

Credential breaches and stealer logs

Checks whether the login page tied to a leaked employee credential or stealer-log record is live and accepting input.

Xvigil

API platforms

Tests publicly shared API collections, such as Postman workspaces, to confirm whether exposed endpoints return your data.

Xvigil

Web App Scanner

Checks the asset is reachable, then records TLS, headers, technology and versions, including exposure to CVEs such as PHP2Shell and Node2Shell.

Bevigil

Network Scanner

Records open ports, banners, certificates and versions, and flags service and kernel CVEs such as CVE-2026-98432.

Bevigil

API Scanner

Records endpoints, methods, status and sanitised responses, with auth context.

Bevigil

AI Events

Checks exposed AI infrastructure, such as MCP servers, LLM endpoints and leaked AI credentials, and whether it is reachable without authentication.

AIvigil

More products and modules

New products and modules follow the same four steps as they are added.

Extensible
1

Observation

What was found, and where

2

Validation

Checks that the signal is still live

3

Context

Asset, technology and impact

4

Recommendation

Verdict, rationale, next step

Value by role

See the analyst hours your team gets back.

Track findings triaged, what the agents found and the analyst hours saved, by module. Then hand the same record to the tools your SOC already runs.

What an analyst opens
Actionable
XVigilAgent triaged
Stealer-log record
urlportal.acme.com/login
secret••••••••••
Employee credentials still work
Login page is live and accepts input.
3 accounts match current employees on your domain.
Recommended next step: reset the passwords and enforce MFA.
Download PDF report
Finding receivedXVigil flags a stealer-log record
Automated
Credential and form agentLogin page checked safely
Identity contextAccounts matched to your domain
Decision AgentVerdict, rationale and next step
Automated
Analyst reviews and decidesAccept or override the verdict
Your team
Analyst Time Saved, by module
Fake URLs and phishingXVIGIL
High
Credential breachesXVIGIL
High
Code repositoriesXVIGIL
Medium
API platformsXVIGIL
Medium
Web App ScannerBEVIGIL
High
Network ScannerBEVIGIL
Medium
API ScannerBEVIGIL
Medium
AI EventsAIVIGIL
Medium
Verdict mix
Actionable
Potential
Non-actionable

Pick a date range to see the verdicts and hours behind each module.

Nexus AICorrelates related findings across products and quantifies the risk for reporting.
Investigation record
Exports as PDF
Record contents
EvidenceRecord, login check
Steps4 of 4 recorded
RationaleWritten summary
ConfidenceHigh
Next actionReset and enforce MFA
Analyst decisionAccepted
  • Signal recordedStealer-log record captured with its source.
  • Live checks completedLogin page live and accepting input.
  • Impact assessed3 accounts matched to current employees.
  • Verdict issuedActionable: reset the passwords and enforce MFA.
  • Decision loggedVerdict accepted by the reviewing analyst.
What your tools receive
Triage once, reuse everywhere
Triage recordActionable
findingEmployee credentials in a stealer log
verdictActionable · confidence high
impact3 employee logins still work
evidenceRecord, login check, domain match
rationaleLogin page live, accounts current
next_actionReset passwords, enforce MFA
decisionAwaiting analyst approval
Delivered throughCloudSEK APIsIntegrationsCloudSEK MCPReports
SOAR playbookOpens the ticket with the action already filled in: reset 3 accounts, enforce MFA, evidence linked. No enrichment step.
Remediation ticket
SIEMReceives the verdict and impact with the alert, so the alert carries its severity and context instead of a raw indicator.
Enriched alert
Your AI assistantAnswers "what do we do about this" from the record, over the CloudSEK MCP, instead of re-reading screenshots and code and spending tokens on checks already done.
Answer with evidence

Tools act faster because the evidence already exists. The approval gate does not move: remediation still waits for your analyst.

Governance

Your team decides what happens after the agents investigate.

Checks that change nothing

Agents run non-destructive checks. Active validation, such as testing whether a leaked key still works, runs in an isolated sandbox.

Every verdict shows its evidence

Each verdict links to the observations, validations, context and rationale behind it, with a recommended next action.

No auto-remediation. No auto-closure.

Nothing changes in your environment without your team's approval. Analysts remediate and close incidents.

A record people and tools can use

Each investigation keeps its evidence, steps, confidence and next action. Analysts read it as a PDF report. Your SOC tools and AI assistants receive the same record through CloudSEK APIs, integrations and the CloudSEK MCP, so remediation starts from evidence that already exists.

Where your analysts sit in the processApproval gate
Step 1Agents investigateNon-destructive checks, in parallel
Step 2Decision Agent recommendsVerdict, rationale, next step
Step 3 · HumanYour analyst reviewsAccept or override Pending
Step 4Your team actsRemediation and closure

Nothing leaves review without an analyst's approval.

FAQ

Frequently Asked Questions

Questions security teams ask first.‍

Does Agents Hub replace analysts?

No. It takes on the repeatable first pass of investigation and evidence gathering, so analysts can focus on judgement, prioritisation and response.

Icon - Elements Webflow Library - BRIX Templates

Can agents make mistakes?

Yes. That is why every verdict comes with its evidence and step-by-step reasoning. Analysts can review and override any verdict, and no remediation happens without their approval.

Icon - Elements Webflow Library - BRIX Templates

Is our data safe?

Agents operate within the CloudSEK platform's existing security boundary. Validation code runs only in sandboxed environments, and agents do not access your production systems or internal infrastructure.

Icon - Elements Webflow Library - BRIX Templates

Which findings get triaged?

Eligibility depends on the product, module and finding type. For example, fake URL triage covers high-severity findings where a screenshot and a live brand site are available for comparison, and AIVigil findings are triaged as AI events. Coverage grows as products and modules are added.

Icon - Elements Webflow Library - BRIX Templates

Does a non-actionable verdict delete the finding?

No. The finding and its evidence stay available, and Agents Hub does not close incidents automatically.

Icon - Elements Webflow Library - BRIX Templates

Can our own SOC tools and AI assistants use the triage output?

Yes. The triage record, with its verdict, evidence, rationale and recommended action, is available through CloudSEK APIs, the Integration Marketplace connectors for SIEM, SOAR, ITSM and messaging tools, the CloudSEK MCP for AI clients, and PDF reports. Your tools start from the record rather than repeating the investigation, and the approval gate stays with your analysts.

Icon - Elements Webflow Library - BRIX Templates

See an investigation from first signal to analyst decision.

Book a demo and we will walk through how Agents Hub handles findings from CloudSEK products, and where your team stays in control.