Sovereign cybersecurity, within everyone's reachSovereign security for allSOC AI on your servers · free edition Request a demoDemo

Enterprise edition

The sovereign SOC AI, with service and support

What belongs to the Enterprise edition is the layer that scales the core: the multi-agent pipeline (Triage, Threat Hunting, Compliance, Remediation), contextual enrichment, ATT&CK coverage, agentic tuning of your rules, direct connectors and output to your SIRP. The three pillars do not change: local sovereignty, explainable AI, human validation.

An alert's journey, in pictures
Ingest, explained verdict, analyst approval, dossier signed twice, then degraded mode. Silent video, on a loop. Illustrated scenes with fictitious data.
Frameworks addressedNIS2DORAISO/IEC 27001MITRE ATT&CKIEC 62443NIST CSFCRAISO/IEC 27090ANSSI

The blind spot

What a triage platform cannot see

Every security platform fires on an alert. A stopped EDR sends none, nor does an unplugged sensor, and an asset nobody ever audited stays green by default. Together these three blocks answer a single question, which assets are exposed, and why, from the telemetry LUAXERR already ingests: nothing more to collect, no agent to deploy.

Vulnerability prioritisation

Your scanner finds forty thousand, your team fixes two hundred a month. LUAXERR does not try to find more: it cross-references your vulnerabilities with asset criticality and the alerts actually observed on them. Exploitation probability (EPSS) and confirmed exploitation (KEV) arrive in a signed offline pack: no estate inventory handed to a third party.

Control coverage

Which assets is nobody watching any more? LUAXERR cross-references your inventory with the telemetry it already ingests and flags silent controls: a machine still reporting through the SIEM, but no longer sending anything from its EDR. And never a hasty conclusion: a silent control is flagged for verification, not declared missing.

Configuration audit

Your hardening reports (Wazuh SCA, OpenSCAP, CIS) return hundreds of red lines, flat. LUAXERR ingests them and says which gaps matter: rule severity, asset criticality, alerts observed on it, how long the gap has stayed open. It opens no privileged session on your equipment. And it will never say “compliant”: the scope actually audited is shown before the gaps, because an asset nobody has looked at is not a healthy asset.

PLCs and connected devices

When an industrial probe feeds LUAXERR, it counts as a control family just like the EDR or the firewall: the analysis can then say which controllers and connected devices nobody watches any more, without ever emitting anything towards the plant floor, where an active scanner is refused outright by operations.

Our design choices

Three subjects we chose to tackle

They do not fit in a comparison grid. They come up the day a source goes quiet, the day an alert arrives through a contractor, the day you need to share a signal without giving anything away. Here is how we handle them.

Source continuity

If an EDR or a cloud SIEM goes quiet, you need to know. LUAXERR watches its own intake paths and tells a silent source apart from a real cutoff. As a fallback, a small pack of signed rules detects on its own: the analyst declares the switch to degraded mode, nothing flips by itself. A cutoff drill simulates losing the model, the vendor, or case exports, so you know before the incident what still holds without them.

Third-party registry

Did an alert come in through a contractor, an integrator, a remote access path? LUAXERR discovers no third party on its own: it attaches recent alerts to the registry you declare yourself. The result says which access path activity came through, not who is at fault: a high volume flags a heavily used path, not a compromised third party.

Signed sharing to MISP

Compose a signal: hypothesis, indicators, a reviewed Sigma rule, ATT&CK techniques, TLP level. Nothing goes out without a preview, a signature and administrator rights, to your local MISP feed. Signals received from a partner run the other way: verified, replayed against your own history, proposed by the Tuning agent, never applied without you.

Example screenThe Enterprise dashboard: blind spots and per-asset posture
The Enterprise dashboard: blind spots and per-asset posture A mock-up of the dashboard. At the top, four figures: silent controls, undetermined assets, assets outside the inventory, checks without a verdict, with the observation window and the age of the intelligence. Below, three assets with their exposure, monitoring, hardening, and what remains unknown about them. luaxerr · enterprise · posture POSTURE ALERTS NIS2 BLIND SPOTS: CONTROL COVERAGE & HARDENING 14-day window · intel 3 d old · last audit 09-08 3Silent controlsthe asset still speaks elsewhere 5Undetermined assetsno telemetry, from any source 12Outside inventoryseen, never declared 7No verdictan unreadable result is not a pass ASSET POSTURE: THE THREE ANGLES TOGETHER ASSETEXPOSUREMONITORINGHARDENINGWHAT WE DO NOT KNOW sql-prod01criticality · essential HIGH · 3 findings · 1 KEVEDR silent4 gaps · critical EDR SILENT · 6 D wks-lab-207criticality · unknown not reportedno silent controlnever audited NOT IN INVENTORY plc-s7-1200criticality · OT · L1 MODERATE · 1 findingOT probe · currentnot audited NEVER AUDITED No composite score: three scales do not average. An asset missing from a column is not healthy, it is unknown to that block. luaxerr · enterprise · posturePOSTUREALERTSNIS2BLIND SPOTS:CONTROL COVERAGE & HARDENING14-day window · intel 3 d oldlast audit 09-083Silent controlsthe asset stillspeaks elsewhere5Undetermined assetsno telemetry,from any source12Outside inventoryseen, never declared7No verdictan unreadable resultis not a passASSET POSTURE:THE THREE ANGLES TOGETHERsql-prod01criticality · essentialEXPOSUREHIGH · 3 findings · 1 KEVMONITORINGEDR silentHARDENING4 gaps · criticalWHAT WE DO NOT KNOWEDR SILENT · 6 Dwks-lab-207criticality · unknownEXPOSUREnot reportedMONITORINGno silent controlHARDENINGnever auditedWHAT WE DO NOT KNOWNOT IN INVENTORYplc-s7-1200criticality · OT · L1EXPOSUREMODERATE · 1 findingMONITORINGOT probe · currentHARDENINGnot auditedWHAT WE DO NOT KNOWNEVER AUDITEDNo composite score: three scales do notaverage. An asset missing from a column isnot healthy, it is unknown to that block.
What the analyst sees: what went quiet, what was never declared, what was never looked at. All of it before the alerts, not after.

The building blocks

What the Enterprise edition adds

Each building block is a concrete operational gain.

Multi-agent pipeline

Each alert flows through a chain of sovereign agents (Triage → Threat Hunting → Compliance → Remediation), under guardrails: a closed action catalogue, and a human in the loop on every heavy action (Human-in-the-Loop).

Compliance & Impact: GDPR, NIS2, DORA, AI Act, NERC CIP

On an incident, the agent derives the obligations of the regime that applies, each with its deadline, authority and legal basis: GDPR, NIS2, DORA, the AI Act, and NERC CIP for a North American power grid. Every regime also shows what the engine simplifies: a wrong compliance deadline is worse than no tool at all.

Agentic tuning

The agent reasons about the “why” of the noise and proposes nuanced multi-condition exclusions, always backtested before you deploy them.

Agentic investigation

Beyond open correlation: an incident narrative, a campaign assessment and a recommended action, backed by auditable reasoning.

Agent memory

The agents remember. Every incident feeds a local memory, episodic (have we seen this IP, this host?) and semantic (a similar campaign?). Facing an alert, LUAXERR recalls linked past cases and flags recurrence. 100% on your side, nothing leaves.

Direct connectors

EDR, IDS·IPS and NGFW directly (CrowdStrike, SentinelOne, Microsoft Defender, Suricata, Palo Alto Networks, Fortinet FortiGate), without relying on an intermediate SIEM.

Critical e-mail alert

Outside working hours, a real attack waits for a human to open the dashboard. Not anymore: e-mail triggered on critical detection, verified TLS, on-call addresses protected.

Signed SOAR output

The case is sent to your SOAR or ticketing in signed form. You integrate the decision into your existing response chain, no copy-paste.

SSO · RBAC · audit

OIDC/LDAP SSO, role-based access control, audit log and compliance reports, for operation at team scale, on a dedicated instance.

Contextual enrichment

Before judging, LUAXERR situates the alert: asset criticality, trusted asset, service account, recent frequency of the rule and entity. It's the context an analyst would have at hand, from your local sources.

ATT&CK coverage

A heatmap of the 14 MITRE ATT&CK tactics projected onto what your estate actually generates: techniques seen per tactic and blind spots at a glance.

TheHive output

On a true positive, LUAXERR creates a case in TheHive (the European open-source SIRP), with the reasoning and observables, never on a false positive. The end-to-end sovereign stack: Wazuh → LUAXERR → TheHive.

Reliability

The harness

More than just an AI model, LUAXERR is the operational backbone of your agents: context mastery, memory management, smooth delegation and sustained human control. A complete framing layer that runs entirely on your own infrastructure.

Context

The ephemeral (an alert's enrichment) and the durable (a case's state) are explicitly separated, computed once and shared across every pipeline stage.

Memory

Past incidents are retained and recalled, by exact entity and by similarity of meaning. The agent “remembers” nearby campaigns as it investigates.

Delegation

A chain of specialised agents (Triage, Threat Hunting, Compliance, Remediation) hands off under guardrails, rather than a single agent that claims to do it all.

Human control

Closed action catalogue, bounded LLM budget, deterministic fallback, human in the loop, audit log. The model informs, it never decides alone.

Our stance: never hand a SOC to the fragile memory of an “autonomous” agent. Determinism and a human where it matters, serious memory for the rest. Reliability first, sovereignty always.

Compute tier

What the multi-agent pipeline requires

The multi-agent pipeline (model-reasoned tuning and investigation) requires more compute, ideally a graphics card, where the free edition's triage is content with a standard processor. LUAXERR runs on the hardware you already have, and sizing is scoped with you at deployment.

Proof for your auditor

A signed dossier, verifiable without us

0bytes

Network egress, verifiable at your firewall

320cases

Control set the model was never trained on. Public measurement method

24h

NIS2 early warning tracked to its deadline, authority named

Hybrid

Signed audit dossier, Ed25519 and post-quantum (ML-DSA-65), verifiable offline without us

SHA-256

Prompt fingerprint written into every verdict, which replays

DiagramThe auditability dossier, and who verifies it
The auditability dossier, and who verifies it A period dossier lists the models, prompt fingerprints, decisions and human control, plus its reservations. It is signed twice, with Ed25519 and with ML-DSA-65, a post-quantum signature. The auditor verifies it offline with the public key, with no return to the vendor. AUDITABILITY DOSSIER · PERIOD SEPT 01-30 · EXAMPLE modelmistral-7b · q4 · sha256 a91f…3c prompt fingerprintc41e…9a (v4) decisions1,284 · human control 96% accessfull log, 3 accounts reservations“traceable, not bit-for-bit reproducible” SIGNEDTWICE public key Your auditorverifies offline, alone LUAXERRno return to us AUDITABILITY DOSSIERPERIOD SEPT 01-30SIGNEDTWICEmodelmistral-7b · q4 · sha256 a91f…3cprompt fingerprintc41e…9a (v4)decisions1,284 · human control 96%accessfull log, 3 accountsreservations“traceable, not bit-for-bit reproducible”public keyYour auditorverifies offline, aloneLUAXERRno return to us
The figures in this diagram are an example. The dossier lists what it does not establish. A document that only states what it proves leaves the auditor to find the holes alone, and loses its credibility exactly when it matters.

Service & support

Sovereign software, a human contact

Deploying an AI into a security chain is not just installing a binary. The Enterprise edition comes with a contractual service: someone answers, the solution stays up to date against threats, and your analysts are supported to autonomy.

Technical support

A dedicated channel for your administrators and analysts, with SLA-guaranteed response times. We diagnose with you, on your deployment, not a ticket that gets lost.

Updates & patches

Patches, security updates and new versions of the specialised model, delivered for the whole subscription period. Your detection does not fall behind threats.

Deployment support

Integration with your SIEM/EDR, fine-tuning on your analyst feedback, runbooks and configurations tailored to your context. We put the tool into production with you, not for you.

Analyst training

Your teams learn to read verdicts, validate a backtested tuning and leverage agentic investigation, to get full value from the tool, without lock-in.

Periodic review

A regular review of verdict quality, the noisiest rules and the tuning gains achieved. The tool sharpens with your real usage.

Sovereign contact

A French vendor, reachable, who knows the solution end to end. No offshored support on security software that no one masters in-house.

TableThree support levels

Three support tiers, aligned with the criticality of your operation. Exact times are set in the quote and contract.

Tier Standard Business Critical
Coverage Business hours Extended business hours Extended on-call
Response (blocking incident) Next business day A few business hours Short, priority time
Channel E-mail / ticket Priority e-mail / ticket Dedicated contact
Security updates Included Included Included
Support & training Optional Onboarding session Continuous support
Periodic review Not included Quarterly Monthly

The tiers, times and scopes above are indicative and framed contractually (see the support and maintenance terms of the T&Cs). Support does not include custom development, governed by a separate engagement.

The trial

A free 30-day trial, on your workstation

To see the Enterprise edition run on your side, with no purchase and before any commercial discussion.

  • What it grants. For 30 days, on one workstation and outside production, use of the Enterprise edition features.
  • What it does not grant. Production use, operation on behalf of a third party, redistribution, or use of these features beyond the term.
  • How it is verified. By a licence key signed by the publisher, which carries its expiry date and is verified offline: no online activation, no outbound call.
  • At expiry. The Enterprise features close. The free edition keeps running with no time limit, your data stays on your infrastructure, nothing is erased or locked.

Request a trial

Contact

Talk about a deployment

A scope to frame, a support tier to choose, a sovereignty constraint? Let's discuss directly, with someone who knows the solution.

Contact us

This site uses cookies. Technical cookies are required for the site to work and stay on at all times. Functional and audience-measurement cookies wait for your consent.

Customise