Aelix Core · The platform layer
The secure foundation
the rest of your utility stack runs on.
Aelix Core is the secure foundation and the agent engine the rest of the Aelix suite runs on. Every product's AI agents run on Core's built-in agent and workflow engine, governed by the same audit trail. It is also the security and compliance backbone underneath: explainable IT-to-OT threat scoring that fails closed, unified logical and physical offboarding, and NERC CIP evidence with a sealed audit pack. Single-tenant, deploy-anywhere on-premises or in your private cloud.
Agentic AI · The agent platform · Domain agents available now
The agent platform your whole utility stack runs on. With agents of its own.
Core is not just where your data and security live. It is where your AI agents live. Aelix Core's built-in agent and workflow engine is the governed runtime that every product's domain agents are deployed on, and it ships with a low-code agent builder so you can build your own. These are domain agents purpose-built for utility work, not a generic chatbot. They watch your systems, reason over live data, draft the next action, and execute it on approval. Every action scoped by role and written to the append-only audit trail.
The platform layer, run the suite's agents or build your own.
One governed runtime.
GridIQ, AssetGuard, CX, and GreenTrack agents all run on Core's built-in agent and workflow engine. Single-tenant, role-scoped, inside your deployment. Data never leaves your boundary.
Low-code Agent Builder.
Build, govern, and deploy your own AI agents and automations on the same engine, with run history and the same audit spine every Aelix agent uses.
Core's own agents (available now)
Threat Triage Agent.
Triages IT and OT alerts, scores them, and recommends or, on approval, executes containment.
Offboarding Orchestration Agent.
Runs the unified logical + physical offboarding workflow end to end across every connected system.
Compliance Evidence Agent.
Continuously gathers NERC CIP evidence and assembles sealed audit packs.
Access Review Agent.
Drives periodic access certifications and flags drift.
Agents do the work. People keep the authority.
Consequential actions (containment, offboarding execution) require a human to approve before they commit. Agents monitor, reason, draft, and recommend. The analyst or operator signs off. Safety gates are never overridden. Every agent action, drafted or executed, writes to Core's append-only audit trail: governed, attributable, and reversible.
The problem
Breaches, audit findings, and orphaned access all live in the same place. The seam.
Every utility runs two worlds that were never designed to meet. A modern, internet-connected IT estate: corporate email, Microsoft 365, cloud workloads, software-as-a-service vendors. And a legacy, high-consequence operational technology (OT) and supervisory control and data acquisition (SCADA) estate that controls physical grid assets: substations, breakers, microgrids. Attackers use the soft IT surface as a beachhead to move toward OT, where a successful intrusion is not a data breach. It is a loss of power to homes, hospitals, and industry.
Three seams sit between those worlds, and no single tool owns any of them.
The IT-to-OT boundary.
Lateral movement from a compromised corporate inbox toward the control network is the highest-consequence undefended path in most utilities, and the hardest to score, explain, and stop without blocking legitimate operations.
The gap between a terminated password and a live substation keycard.
Manual offboarding routinely revokes the login but misses the physical badge, or vice versa. A terminated insider who still holds substation access is both a security failure and a safety failure.
The gap between an auditor's question and the evidence.
NERC Critical Infrastructure Protection (CIP) requires evidence collected on prescribed cadences and retained for years. Stitched together from disjoint security tools, spreadsheets, and manual checklists, that evidence is a standing liability, and producing it consumes security staff for months.
The cost of doing nothing is asymmetric: a missed evidence cadence is a fine. A successful IT-to-OT pivot is a grid event. Aelix Core is built to close all three seams on one append-only record.
What Core does
One platform to secure, comply, and recover.
Explainable IT-to-OT threat scoring that fails closed
Core scores network flows across the IT-to-OT boundary with a machine-learning (ML) model and shows analysts why: the verdict, a confidence score, and per-feature attribution. Not a black box they cannot defend to a regulator or a boss.
Flow telemetry is scored by a dedicated inference service reached only over mutual TLS. If the model is unavailable, Core falls back to a Suricata rule engine and signals "model unavailable" in the interface. It never silently allows a flow. Fail-closed is the design, not a configuration option.
Unified logical and physical offboarding
When an employee is terminated, Core revokes software access and physical access in lockstep: Microsoft 365, Active Directory, virtual private network (VPN), SCADA, and ServiceNow alongside substation keycards and badge systems. No one keeps a key after they lose their login. This combined logical-plus-physical revocation is the differentiated wedge. Most tools touch only one side of the seam.
A termination orchestrator drives badge connectors (designed to connect to HID, Genetec, Lenel, Honeywell, and Brivo) and software connectors (Microsoft 365, Active Directory / Okta, Amazon Web Services, ServiceNow, Splunk, and more). Each revocation runs as an idempotent child job with retry and exponential backoff, streams live per-step progress, and writes one audit row per step. The operator types the employee identifier to confirm before the wizard runs.
NERC CIP evidence with a sealed audit pack
Core treats CIP evidence as an automated workflow, not a quarterly spreadsheet scramble. It maintains a CIP-002 through CIP-013 control catalog with per-control status, gathers evidence on each control's required cadence, and exports a sealed audit pack ready for federal submission.
Scheduled collectors are designed to gather evidence from sources including CrowdStrike, Splunk, Windows Server Update Services, Active Directory, Nessus, access-control, learning-management, and patch-management systems. Evidence is stored with a 7-year lifecycle per CIP-008, and the audit-pack export bundles a zip, a manifest, and a cryptographic hash so the package is tamper-evident.
Append-only audit trail as the system of record
Every state-changing action in Core (an analyst resolving a threat, an employee offboarded, evidence collected, a role changed) writes one immutable row. That trail is the compliance system of record, not a scattered set of logs.
Each state change is recorded through a single audit service. The audit table is enforced append-only at the database level: the application connects as a non-superuser, row-level-security-bound role, so rows cannot be edited or deleted at runtime. Reads support filtering, cursor pagination, and newline-delimited JSON export, and every row carries a request correlation identifier.
The built-in agent and workflow engine the suite runs on
Core is the foundation layer. Aelix Core's built-in agent and workflow engine plus the shared security, identity, and audit services that the other four Aelix products (GridIQ, AssetGuard, CX, and GreenTrack) build on. Every product's AI agents are deployed and governed here. Build your own AI agents and trigger-condition-action automations with the low-code agent builder, complete with run history, and reuse the same role-based access control, multi-factor authentication, notifications, and audit spine everywhere.
The agent builder composes trigger to condition to branched-action flows, and multi-step agent logic, over a local runtime, with every action written to the append-only audit trail. Identity is handled by Argon2id local authentication with time-based one-time-password multi-factor authentication enforced for privileged roles, and single sign-on / cross-domain identity provisioning (SCIM) designed to connect to Okta, Active Directory, and Ping Identity.
Single-tenant, deploy-anywhere
One deployment per utility. Your data and your stack are physically isolated to your environment. Isolation is the feature, not the limitation. Run Core on-premises or in your private cloud. There is no shared multi-tenant infrastructure and no shared-cloud dependency.
Core ships as a containerized stack you deploy on-premises first and migrate to Amazon Web Services, Google Cloud, or Azure when you choose. Single-tenancy is an architectural invariant, which also makes NERC CIP scoping simpler because the boundary is your deployment.
Why Aelix Core
The platform layer, and the security backbone underneath it.
It is a foundation and the agent engine, not a point tool.
Core is the secure architecture and built-in agent and workflow engine the rest of the Aelix suite sits on. GridIQ (control room), AssetGuard (asset health), CX (revenue), and GreenTrack (renewables and ESG reporting) inherit the same identity, role-based access control, notifications, and append-only audit trail from Core. Every domain AI agent they ship runs on Core's governed agent runtime. You adopt one security, compliance, and agent backbone once, and every product you add reuses it.
It closes the seam, not just one side of it.
The differentiated capability is unified logical and physical offboarding: revoking a login and a substation keycard in the same orchestrated, fully audited action. That combination is the gap most security stacks leave open.
It is honest by default.
Core fails closed and never silently allows a flow. Verdicts are explainable (confidence plus feature attribution) so an analyst can defend a decision to a regulator. Every external integration runs in a clearly labeled simulated mode until you wire your real provider, and the audit record is real in both modes. We do not publish a 'five nines' availability figure we have not soak-tested, and we do not badge certifications we do not hold.
It keeps a human in the loop.
Core surfaces explainable scores and orchestrates revocations and evidence collection. The analyst acknowledges and resolves, the operator confirms, and the disaster-recovery runbook pauses for signoff. The platform makes the defensible action fast. It does not take high-consequence actions silently on your behalf.
What we can prove, and what we are targeting
Mechanisms you can verify. Targets we are honest about.
Aelix Core is production-shaped, not yet production-validated. The figures below are engineering targets and design mechanisms, clearly labeled as such. Not customer outcomes. We will publish measured baselines from our first design-partner deployments as they are established.
Targets (engineering goals, not measured outcomes)
- TargetOffboarding completeness: revoke all 10+ logical and physical access items in under 5 minutes, with 100% of revocations audit-logged.
- TargetThreat-scoring latency: ML inference under 150 milliseconds at the 99th percentile on the CPU/ONNX path. (Requires load validation.)
- Designed-in mechanismFail-closed coverage: 100% of model-down events fall back to the rule engine and never silently allow.
- TargetEvidence automation: automated evidence collection on cadence for at least 90% of in-scope CIP controls.
- Designed-in mechanismAudit completeness: 100% of state-changing actions write an audit row.
Mechanisms you can verify in a demo
- A unified offboarding run revoking 10+ logical and physical items live, every step audit-logged. (Demo data.)
- A sealed NERC CIP audit-pack export: zip plus manifest plus cryptographic hash. (Demo data.)
- An explainable threat verdict showing the model, confidence, and feature attribution. (Demo data.)
Security and compliance posture
Designed for the standards your auditors hold you to.
Compliance framing (honest)
Designed to NERC CIP standards. Append-only audit trail, role-based access control, scheduled evidence collectors, and 7-year CIP-008 evidence retention engineered to support your CIP scoping. CIP-003-9 is enforceable 2026-04-01. Aelix Core is not a vendor-held certification. Compliance is scoped per utility and per deployment, and Core is built to support that scoping.
Architecture facts (true today)
Single-tenant isolation.
One deployment per utility. Your data and stack are physically isolated to your environment.
Deploy-anywhere.
On-premises first, portable to your private cloud. No shared multi-tenant infrastructure.
Append-only audit trail.
Database-enforced via a non-superuser, row-level-security-bound runtime role.
Encryption in transit and at rest.
Mutual TLS for service-to-service traffic, including the ML inference path. Encryption at rest for stored data and evidence.
Least privilege.
The runtime connects as a non-superuser role. The superuser is reserved for migration and forensics only.
Role-based access control and multi-factor authentication.
Time-based one-time-password MFA enforced for privileged roles. Single sign-on and SCIM provisioning designed to connect to Okta, Active Directory, and Ping Identity.
Hardened edge.
TLS termination, content-security-policy, HTTP strict transport security, frame and referrer protections, request-size limits, and per-IP rate limiting.
Secrets management.
Secrets sourced through a secret-provider adapter (HashiCorp Vault), with mutual-TLS certificate and key rotation by design.
A note on availability. We deploy single-tenant in your environment and design availability around your operational requirements. We do not publish a shared-cloud uptime figure we have not soak-tested.
See it for yourself
See Core close the seam.
Walk through a unified offboarding, a sealed NERC CIP audit pack, and an explainable IT-to-OT threat verdict, deployed the way you would run it: single-tenant, in your environment.