MDx OS

The System.

The operating system for AI-native work.

Not a platform. Not a framework. An operating system. The layer underneath where models, agents, context, and governance converge — so every app on top feels like one system.

Applications
Code
Twin
Message
Stella
Pulse
OS Service APIs
MDx OS
Model Router
Context Engine
Orchestrator
Governance
Provider Adapters
Memory
Audit Trail
Stream Fabric
Why "OS"

The same thing. For AI.

A traditional OS manages CPU, memory, storage, and processes. MDx OS manages models, context, knowledge, and agents. The mapping isn't hand-wavy — it's structural.

Traditional OSMDx OS
CPU (compute)Models (AI compute)
MemoryContext (working memory for agents)
StorageKnowledge (persistent org knowledge)
Process lifecycleAgent lifecycle
Process schedulingOrchestration + intelligent routing
I/O interfacesChat, voice, API, WebSocket
Security / permissionsGuardrails, governance, audit
Device driversProvider adapters (Claude, Grok, Gemini, OpenAI)
The Problem

Every vendor is building their own AI silo.

Each one locks your context, your memory, and your model choices inside their walled garden.

Salesforce Agentforce
Agents that only work inside Salesforce. Memory that doesn't port. Context locked to Einstein.
Microsoft Copilot
Locked to M365 and Microsoft Graph. Your organizational intelligence trapped in one ecosystem.
ServiceNow AI
Workflows that only run on ServiceNow. Skills that don't transfer. No cross-system context.
Pega GenAI
AI embedded in Pega decisioning. Your models, your rules — but only inside their runtime.
MDx OS
One routing engine. One context layer. One agent lifecycle manager. One audit trail. The vendors provide application capabilities. You own the intelligence layer.
Architecture

Four layers. Fifteen components.

Every component serves any application — not just MDx. The Model Router doesn't know what's calling it. The Context Engine manages profiles for any consumer.

01 Foundation
The primitives everything else depends on. Model abstraction, context management, persistent knowledge, cost tracking.
Model Router
Context Engine
Knowledge Architecture
Memory
Provider Adapters
02 Agents
Agent lifecycle, routing, delegation, parallel execution, skills framework, tool integration.
Orchestrator
Specialist Agents
Skills Framework
Tool Integration
03 Governance
Every AI decision is auditable. Input/output guardrails, RBAC, human-in-the-loop approval, chain-hashed audit trail.
Guardrails
HITL Protocols
Audit Trail
Security & RBAC
04 Evolution
The OS gets smarter over time. Learning loops that improve routing, agent quality, and knowledge — automatically.
Learning Loops
Federation (MCP)
Metrics & Observability
Ecosystem

Five apps. One OS.

Each app was built to solve a different problem. Each one runs on the same foundation — shared auth, shared context, shared governance.

C
MDx Code
AI engineering manager. Orchestrates Claude, Codex, and Gemini with adversarial review, audit trails, and cost tracking.
Phases 7-14
T
MDx Twin
Cognitive twin. Five specialist agents that carry your expertise, judgment, and voice. Your strategic mirror.
Phases 1-6
M
MDx Message
Real-time communication where humans and agents talk. Rust relay, typed messages, observable orchestration.
Phase 17
E
Stella
Agentic wellness benefits coach. Three domain agents — Health, Wealth, Protection — with signal detection and confidence scoring.
Phase 15
S
Pulse
Wellness intelligence dashboard. Biometric signals, Apple Health integration, benefits activation engine.
Phase 16
P
SDLC Pipeline
AI-native software delivery. //plan, //run, //close. Expert panel, governance, gap triage — intent to production autonomously.
Phase 18
Maturity Model

Five levels. From copilot to autonomous.

MDx OS is designed to take organizations from L1 (AI as autocomplete) to L5 (AI as autonomous workforce). Most enterprises are stuck at L2.

L1 Assisted AI as autocomplete. Code suggestions, chat summaries. No real agency. Commodity
L2 Augmented Multi-model routing, context injection, tool use. Human starts every action. Complete
L3 Agentic Agents delegate, hand off, run in parallel. Human-in-the-loop at decision points. Current
L4 Autonomous End-to-end workflows without human initiation. Policy-governed, audit-trailed. Building
L5 Orchestrated Cross-system agent federation. MDx OS as the connective tissue across the enterprise. Architected
The Arc

Built from the inside out.

Apple didn't design macOS in a vacuum and then figure out what to run on it. They built apps — and the OS grew underneath them. The apps drove OS development. The OS enabled more apps.

MDx follows the same pattern. Twin was app #1. The orchestration, routing, and context underneath it became the OS. Code was app #2. Stella, Pulse, Message followed. Each one proved the OS was real — not just good architecture for one app.

When five fundamentally different apps run on the same infrastructure — you actually have an OS.
Build Timeline
Ph 1-6 Twin OS patterns emerge
Ph 7-14 Code OS services formalized
Ph 15-16 Stella + Pulse OS proven
Ph 17 Message Stream Fabric primitive
Ph 18 SDLC Pipeline Autonomous delivery (current)
5
Apps on OS
15
OS Components
326K+
Lines of Code
3,162
Tests Passing
135
Migrations
139
Sessions Built
19
Phases Complete

Explore the ecosystem.

Every app runs on MDx OS. Pick one and see what it does.

MDx Code → MDx Twin → MDx Message → Back to MDx →
Read the full thesis: MDx OS — The Operating System for the AI Era