Cortex is a sovereign, self-hosted AI control plane that connects to the enterprise systems you already run — Maximo, SAP, Salesforce, Snowflake, your databases — and replaces their interfaces with a single natural-language workspace. Every agent action is cryptographically signed and Merkle-chained. Nothing ever leaves your infrastructure.
Most "AI for the enterprise" platforms either route your data to a vendor cloud, bolt a chatbot onto existing screens, or hand you a governance dashboard with nothing underneath it. Cortex was built around three constraints that don't compromise.
Cortex is a single Rust binary plus PostgreSQL with pgvector — nothing else. It runs on your server, validates its license offline via Ed25519-signed JWT, and never makes an outbound call. Defence, banking, and energy customers can run it fully air-gapped with updates delivered on physical media.
Every agent action — every query, every write, every dashboard generated — produces a BLAKE3-hashed, Ed25519-signed TraceCaps capsule, chained into a Merkle tree and anchored to IETF SCITT. The result is an IETF AAT-compliant audit trail that satisfies EU AI Act Art. 12 and NERC CIP-015-1 out of the box.
Cortex doesn't ask anyone to switch tools. It observes how your teams already work inside Maximo, SAP, or Salesforce, mirrors the underlying data in real time, and gradually generates a faster, accessible interface behind a Strangler Fig façade — with instant rollback to the legacy app if anything doesn't match.
Cortex migrates a workload from legacy software to a Cortex-native, accessible dashboard through a fixed sequence. Each phase is independently reversible, and each one produces its own cryptographic record.
A lightweight browser extension and ObserverAgent record field-level interactions inside your existing applications, building decision traces without changing a single workflow.
Once a field is observed often enough, the Mirror Engine subscribes to the source database's change stream — column-level, Kafka-free — and replicates it into TraceDB with sub-100ms p95 latency at 250M+ events/week.
The Absorption Engine promotes mirrored fields into agent-safe, copy-on-write branches, gated by a write-approval step — so Cortex can act on the data without ever risking the source of truth.
The Genesis Engine converts absorbed fields and observed workflows directly into WCAG 2.2 AA A2UI dashboards — behaviourally equivalent to the screens your teams already know, just faster and accessible.
Live traffic is progressively routed to the new Cortex panels. An absorption-score dashboard shows leadership exactly how much of each system has been replaced, with instant hybrid rollback to the legacy app at the same workflow step if needed.
The Retirement Engine captures full context, replays it for functional equivalence, and signs a cryptographic decommissioning certificate — your evidence that the legacy system can be switched off.
Cortex assumes zero implicit trust between any two components. Every MCP request — from a user, an agent, or another system — passes through the full stack before it can touch a connector.
A single administrator action halts every agent in the system, instantly and verifiably — even with no network connectivity. Recovery is just as immediate.
Built on a hardware-token, behavioural-baseline and network-heartbeat model — works identically whether you're online or fully air-gapped.
Cortex ships with MCP connectors for the systems most enterprises already run, with new connectors added continuously. No vendor-specific runtime dependencies are required for backup parsing.
Every tier runs the same binary and the same provenance engine. The only difference is where the infrastructure — and the keys — live.
Cortex's self-validation suite runs twelve independent experiments — security, semantic routing, provenance integrity, mobile inference and more — and produces a signed, reproducible AnalysisReport.
Cortex's deployment model is validated against Dell PowerEdge XE servers with NVIDIA NemoClaw, producing a due-diligence report and deployment blueprint for direct submission through the Dell AI Ecosystem Program.
View validation reportSystems integrators, MSPs and compliance consultancies can deliver Cortex deployments using the Implementation Manual and a structured Phase 0–6 launch plan. Partner enablement includes access to the Knowledge Snap library and connector SDK.
Apply to partnerBecause Cortex replaces software rather than adding to it, pricing is structured around infrastructure footprint and connector count — not per-user fees that punish adoption. Every plan includes the full Security Fortress and provenance engine.
Documentation is versioned alongside every release and ships inside the air-gap bundle, so your teams never need external access to read it.
Single static Rust binary (≤10MB), Docker image, or a complete offline .tar.gz bundle with binary, config, Knowledge Snap and migrations.
Download latest release →How to use the Interface of One — natural-language queries, dashboards, the command bar, and the Provenance Explorer for compliance teams.
Read the user manual →The full Phase 0–6 launch runbook: install, connector configuration, the Absorption Pipeline, validation suite, and air-gap deployment procedures.
Read the implementation manual →Explore a hosted Cortex instance pre-loaded with a Knowledge Snap for your industry — energy & utilities, banking, or general enterprise.
Launch the demo →Apply for implementation-partner enablement, including the connector SDK, Knowledge Snap library, and joint validation reports.
Apply to the partner programme →Per-node licensing across Pilot, Sovereign Enterprise and Air-Gapped tiers, with a guided consultation to size your deployment.
View pricing →Talk to the team building Cortex about a 90-day pilot on one workload — fully reversible, fully audited, and run entirely inside your own infrastructure.