Modular Coaching Intelligence

Intelligent agents.
Better bodies.


A next-generation coaching platform that separates domain expertise from execution flow. Specialized agents collaborate through composable workflows to deliver safe, adaptive, personalized guidance — across onboarding, program design, progress tracking, and coordination.

6
Agent roles
5
Workflows
3
Patterns
Extensible
Core benefit
Safety-first by design
Architecture
Composable, reusable agents
Orchestration
Sequential, parallel & dynamic
01

Agent roles

Ac
admin_agent

Intake & Admin

Captures member data, onboarding details, and administrative context to build a complete member profile.

Co
coaching_agent

Exercise Coaching

Generates exercise plans, training recommendations, and coaching insights tailored to individual capacity.

Nu
nutrition_agent

Nutrition Guidance

Creates meal guidance and dietary recommendations precisely aligned to each member's fitness goals.

Sa
safety_agent

Safety Validation

Validates load, contraindications, and risk factors before any plan is delivered to a member.

Pr
progress_agent

Progress Tracking

Collects progress updates and tracks outcomes longitudinally, adapting plans as members evolve.

Cx
coordinator_agent

Dynamic Orchestration

Acts as manager — choosing which specialists to invoke, in what order, and when to combine their outputs.

02

Workflow patterns

Sequential

Onboarding workflow

Strictly ordered agent collaboration — intake first, then safety review, then program generation. Ideal for new member onboarding where each step depends on the last.

admin_agent safety_agent coaching_agent nutrition_agent
Concurrent

Data collection workflow

Parallel agent execution — agents run simultaneously for efficiency. Best suited for data collection and review tasks where results are independent.

progress_agent admin_agent
Manager-driven

Coordinator workflow

Dynamic orchestration — the coordinator agent decides which specialists to invoke and in what order, adapting to the complexity of each member's situation.

coordinator_agent all specialists
Sequential

Program planning workflow

Combines safety review, exercise coaching, and nutrition guidance into a unified, consistent plan ready for delivery.

safety_agent coaching_agent nutrition_agent
Sequential

Check-in workflow

Member check-ins that feed directly into progress tracking, closing the loop between plan delivery and outcomes.

admin_agent progress_agent
03

System architecture

Intake
Member data captured and validated by admin_agent
Safety review
Contraindications and risk flagged before planning begins
Plan generation
Coaching and nutrition agents produce a unified program
Orchestration
Coordinator adapts flow dynamically across workflows
04

Project impact

01
Accelerates product development

Domain logic can be built and extended independently by separate teams without coupling concerns.

02
Builds member trust

Explicit safety and validation stages give auditable assurance that no program skips a risk check.

03
Enables future expansion

The composable architecture extends naturally to more advanced coaching scenarios and third-party integrations.

04
Supports deep personalization

Agent behavior can be mixed and matched across different workflows to suit any member profile or goal.

05

Developer guide

Built on the Microsoft Agent Framework. Configure Azure OpenAI credentials and run the onboarding demo. Agents and workflows update independently.

AZURE_OPENAI_ENDPOINT=your_endpoint
AZURE_OPENAI_API_KEY=your_api_key
AZURE_OPENAI_CHAT_DEPLOYMENT_NAME=your_deployment

python main.py
Framework

Powered by the Microsoft Agent Framework — agents, tools, and orchestration are all first-class framework primitives.

Entrypoint

Use main.py for a quick workflow demo that exercises the full agent pipeline.

Extend agents

Add new capabilities under agents/ — each module owns a narrow domain task.

Add workflows

New orchestration patterns go under workflows/ using the existing engine primitives.

06

Next steps

Check-in

Extend checkin_workflow to support ongoing member follow-up beyond the initial program period.

Progress

Add personalized goal tracking features inside progress_agent with longitudinal scoring.

Coordination

Introduce multi-agent decision scoring in coordinator_workflow for smarter specialist selection.