THE PLATFORM

The Wellness Orchestration OS

An operating layer that turns one member intent into a coordinated day of wellbeing — with permissions, boundaries and human accountability at every step.

THE OPERATOR VIEW

The whole centre, in one clear view

Journeys, agents, approvals and outcomes come together in one workspace the team can understand at a glance.

Agentic MikisOperations Console
Live
AGENTIC WELLNESS OS · LIVE VIEW

Good morning, Operations

Your centre is coordinated. Two decisions need you.

Today
Journeys18Active now
Agents12Working
Approvals2Needs you
Completed31without exception
Fictional member
Member journey

Recover from jet lag before Sunday

Mikis, fictional master orchestratorMikisMikis is orchestrating

Today’s plan

Movement sessionTraining AgentReady
Recovery therapyRecovery AgentHeld
Nutrition & diningDining AgentReview
Fictional wellness lead
Your approval

Hydration assessment

Diagnostics recommends a professional review before adding this step.

Wellness lead02:14
ApproveReview
Agent team

4 connected

Training AI agentRecovery AI agentFictional wellness lead+1
Training AI agentTrainingChecking availability
Recovery AI agentRecoveryTreatment held
Dining AI agentDiningAdapting the menu
Safety AI agentSafetyBoundaries verified
Every decision is recordedOpen console
Illustrative product view · no real member data

ARCHITECTURE

Six layers, one path

Intent travels from the member's interface to the physical space and returns as an outcome. Each layer has one clear responsibility.

LAYER 01

Member interface

The member stays in the frontier AI they already trust — ChatGPT, Claude, Gemini, the Greek-language models Meltemi and Kri-Kri, or their personal assistant. No new app to adopt.

LAYER 02

Identity & consent

Every request is bound to verified membership, explicit consent and the exact data scope the member has approved.

LAYER 03

Mikis, the conductor

Interprets intent, decomposes it into tasks, selects agents, holds sequence and timing, and stops where approval is required.

LAYER 04

Branded specialist agents

The centre's private agent family — carrying its name, voice and protocols, not generic bots.

LAYER 05

Systems & humans

Booking, membership, spa, F&B, payments and staff tooling — plus the qualified professionals who remain accountable.

LAYER 06

Physical environment

Spaces, temperatures, equipment and flow — and, in future, Physical AI where it is genuinely useful and safe.

CONNECTIVITY

Permissioned execution, not open access

The difference between an assistant that talks and a system that executes is how permission is granted.

PRIVATE MCP SERVER

The centre keeps control

Capabilities are exposed as tools on a private MCP server the centre owns. The frontier model calls tools — it never gets your database.

LEAST SCOPE

Permissions per task

Each agent receives the narrowest possible scope for a specific task, and only for as long as the task runs.

APPROVAL GATES

Anything risky stops

Clinical, medical, financial or policy-exception steps pass through a human before execution.

TRACEABILITY

Complete record

Every intent, tool call, approval and outcome is logged and can be reviewed.

FLOW

What happens inside one request

  1. 01The member states an intent inside their own AI.
  2. 02Membership, consent and data scope are verified.
  3. 03Mikis decomposes the intent into tasks and selects agents.
  4. 04Agents check availability, protocols and constraints.
  5. 05Anything requiring judgement is routed to a qualified human.
  6. 06The day is confirmed back as one coherent schedule.
  7. 07The outcome is recorded and informs the next visit.