The frontend experience layer of Legion X

The last interface layer
your enterprise will ever need

A single, policy-orchestrated pane of glass that abstracts every legacy and modern system, propagates change instantly, and becomes alive through a per-user AI agent. Every action is governed, audited, and defensible — rendered the way each person needs to see it.

One interface. Every system. Every person.

PanaceaUX is not another frontend framework. It is the interface layer that consumes the Legion X middleware — APIs, MCP servers, and the Policy Engine — and presents one consistent, governed experience across web, desktop, mobile, and operational displays.

01

One pane of glass

A unified client runtime — Flutter/Dart as primary, Uno (.NET) as the backup path for Microsoft-aligned estates — renders every workflow with pixel-consistent behaviour and a single learning curve, no matter the underlying system.

02

Same descriptor, any runtime

Screens are described, not coded. A semantic descriptor served by the platform drives the web edge today and the .NET/Uno client next, with no backend change. The same contract already powers the Alberta Innovates SME-Intake deployment.

03

Governed by design

Every render passes the Legion X Policy Engine first. The interface cannot show data a role may not see, cannot remove an action an integrity rule requires, and stamps every change to the audit ledger.

Every surface defaults to the mode each person needs

Experience Modes are policy, not preference. Each surface queries the user's preferred mode and renders to it by default — the same intent, the same data, the same actions, presented for how that person works. Switch modes below and watch one invoice re-render from a single source.

Simulation — runs entirely in your browser. No AI, no backend, no data leaves this page.

One Semantic Intent Graph, four presentations. The amount, vendor, due date, actions, and policy tags are identical across every mode — only the rendering changes. Two further modes, braille and audibleOnly, follow the same contract.

A graph of intent, decoupled from presentation

The Semantic Intent Graph separates what the interface does from how it looks or where it renders. It is the central nervous system that lets the AI rearrange the interface without ever breaking data integrity, action wiring, or policy — and lets one descriptor serve every mode and every device.

nodes What the interface is made of

  • DataField a value — amount, valve position, weather reading
  • Action something that can be done — approve, open valve, trigger sign
  • View a logical screen or panel
  • Container a layout grouping the AI can rearrange
  • Visualization charts, overlays, correlation bands
  • Trigger an automated or one-click action
  • DeviceSlot a virtual slot for multi-device layout
  • Correlation a synthesized join of live and historical data

edges How they relate

  • contains hierarchical layout
  • triggers one action triggers another
  • visualizes a visualization shows a field or correlation
  • dependsOn a functional dependency that cannot be broken
  • rendersOn a device-slot preference
  • compatibleWithMode links to Experience Modes
  • personalVariantOf a user fork of a golden view
// one node, the approve button { intentId: "invoice.approve", type: "Action", verb: "APPROVE", dependsOn: ["invoice.amount"], policyTags: ["rbac:finance.approver", "audit:always"] }

Speak to the interface. It rewrites itself — within the rules.

An always-present per-user agent rewrites the interface in natural language inside an isolated playground: preview, apply, save a personal variant, or share for approval. Every proposed change passes the middleware Guardrail Validator before it renders. Choose a command and watch the validator decide.

Simulation — the validator logic here is deterministic and runs in your browser. A faithful illustration of the middleware Guardrail Validator, not a live AI.

1 · Parse intent

Select a command above.

2 · Guardrail Validator

Five checks run before any render.

Allowed rewrites mutate the shared interface state and re-render the device mesh below. Refused rewrites never render — the offending check is shown, and a compliant alternative is offered. Zero-trust rendering: every change is a signed, audited transaction.

Dynamic view synthesis across the device mesh

The same intent graph distributes across a confirmed mesh of devices. Ask for a layout and the interface synthesizes it — reassigning slots, building correlated views, wiring new triggers — while preserving every security and audit rule.

Simulation — device assignment is local state. No devices are contacted.

The SCADA example from the field: a valve view with weather-station history, a 30-minute pre-spike alert band, and an automatic road-sign trigger — synthesized from one natural request, governed end to end.

The interface cannot show what policy forbids

Every AI-proposed change is validated by the middleware before a single pixel renders. The Guardrail Validator is a deterministic rules engine — with an optional model second opinion and human-in-the-loop for high-risk changes — and every mutation is a signed, audited transaction with full provenance.

Check 01

Role access

The caller's role must permit every field and action in the change. Denied before execution.

Check 02

Data exposure

No personal or out-of-scope data may be surfaced. Masking and field-level RBAC are enforced at the data layer, not the prompt.

Check 03

Action integrity

A field an action depends on cannot be hidden or removed. Functional wiring is preserved.

Check 04 & 05

Budget & audit

Render stays within the performance budget, and the change is stamped to the append-only audit ledger.

The experience layer over a governed runtime

PanaceaUX does not replace governance — it surfaces it. It sits over the same Legion X middleware that runs governed workflows today, reached through a standard set of MCP servers so any AI host can drive the interface safely, with policy enforced at every call.

intent-graph-mcp

Intent Graph

Query nodes, suggest variants, apply deltas, and validate intent against the live graph.

policy-mcp

Policy & Guardrails

Check RBAC, validate a change, and resolve the modes a user is allowed to use.

device-mesh-mcp

Device Mesh

Discover devices, assign slots, and orchestrate layouts across the confirmed mesh.

experience-mode-mcp

Experience Mode & Runtime

Switch modes, synthesize specialized views, and inspect or preview live changes.

Learn more about the governed runtime at legion.iaxov.com.

From foundation to autonomous evolution

A deliberately phased path: ship the foundation, bring the interface alive, make it intelligent and collaborative, then open the ecosystem. Five tech spikes — guardrail safety, multi-device state sync, rewrite performance, voice-and-agent fusion, and audit-and-compliance — de-risk the hardest parts up front.

Phase 0 · Q3 2026

Foundation

Flutter and Uno runtimes, multi-mode shells, descriptor rendering, the policy engine, FDE tooling, and the realtime voice pipeline.

Phase 1 · Q4 2026

Alive Interface

The personal AI agent and playground, natural-language rewrite with guardrails, and multi-device orchestration across two to three screens.

Phase 2 · Q1–Q2 2027

Intelligent & Collaborative

Dynamic view synthesis, learning and proactive suggestions, variant sharing and approval, and full Uno parity.

Phase 3 · Q3 2027+

Ecosystem & Autonomy

A variant marketplace, AI-to-AI collaboration, continuous optimization, and on-premises or air-gapped deployments.

Forward-deployed engineers get you started; the platform keeps you moving

For larger organizations, IAXOV forward-deployed engineering teams embed to handle migration, build custom modes, and transfer knowledge — then clients graduate to self-service while IAXOV retains platform governance.

A

Embed

FDE teams work alongside your people to migrate systems, author Experience Modes, and synthesize the views your operators actually need.

B

Graduate

Target: 80% of FDE clients reach self-service within six months, co-creating operational excellence with their personal agent.

C

Sustain

IAXOV retains platform governance and value while your teams build, share, and promote their own compliant interface variants.

Bring the living interface
to your enterprise.

Sixty minutes. No slides. A working demonstration of governed, AI-orchestrated interface synthesis applied to your domain. If it does not meet your standards, we have not done our job.