Meet MobiFit: My Fitness App and How It Works

MobiFit is a fitness app that combines AI with multi-agent, local-first functionality, thoughtfully designed UX, groups with an internal social network, and Instagram integration—all so you can train anywhere.

Created by an independent developer, the project was born from the realization that many gym apps are bureaucratic: too many forms, confusing interfaces, and little focus on what matters—getting started with training. A discovery process with students and fitness professionals guided the feature decisions.

The available ready-made workouts were created by human professionals, specifically for the app and for various levels.

Available on: Google Play and App Store | Free
Languages: English, Portuguese, and Spanish

The Experience, Powered by AI

AI isn't an add-on: it's a central part of the architecture and can execute actions within the app, create dynamic content, and use user data to personalize recommendations.

What the AI Does

• Personalized workout plans from natural language.
• Answers to questions (execution, adaptations by level, goal, or equipment).
• Assisted prompt ("Create Workout"): a guided flow that collects preferences and generates the ideal prompt for those who don't know where to start or how to create a good workout prompt. Part of the planning for better user experience.
• Original content: exercise instructions and videos are human-made (not AI-generated), produced by the author with support from professionals. All execution videos were recorded exclusively for this app.

Page AI MobiFit	AI MobiFit Prompt Form	Exercise Block

Agent Architecture

The received message mixes text and structured data in blocks (e.g., exercises, diet, actions). These blocks feed dynamic cards and app actions. Each request is routed to a specific agent.

A request is received by an Edge Function, which executes an AI agent integrated with an LLM model, streaming the response.

For workouts, exercises are grouped by day, with options to save to "My Workouts," where you can create sessions, share with other users, or generate a custom photo for Instagram.

The agent is context-aware: in addition to chat history, it retrieves the user's progress and preferences (when authenticated).

Product Design and UX: Unblocked Journey

The proposition is simple: start training before signing up.

• Unblocked journey: no login, no forms, and no permissions for most initial features.
• Content by level: library for beginner, intermediate, and advanced, with video, image, and text instructions—all material is human-made and copyrighted by the author.
• Meaningful gamification: coins, medals, and rankings that reward consistency.
• Social: groups with internal feed for posts, workouts, and materials; useful space for anamnesis with students/clients.
• Three languages from first access.
• Design principle: reduce effort in the first minutes after installing.

Gamification and Group Challenges

Goals and Challenges

• Goals: Individual with an end date; users check in daily by marking a workout as completed or in the goals menu.
• Challenges: Group-based. Check-in is a post in the group (can include physical activity data and photo); the challenge period can be defined or continuous.

Coins, Medals, and Rankings

• Coins: practically every relevant action earns coins.
• Medals: mark specific achievements.
• Ranking: in challenges, classification uses members' check-ins.

Groups

• Highly configurable with administrators.
• Allow making workouts available to everyone (each member can add to their personal list).
• Members have "profiles."
• Anamnesis sent by members is visible only to administrators (ideal for personal trainers and coaches).
• Currently the only feature that requires authentication.

MobiFit Group	Group Post Feed

Local-First

Data is persisted locally and, when there's internet, synchronized with the cloud (if the user is authenticated). Otherwise, it remains only on the device.

Synchronization uses a simple last-sync strategy: the app stores local date/time and compares with cloud data to decide what prevails.

Result: exercise library and workouts fully usable offline, including edits, progress, and gamification.

Workouts

Hierarchy: Plan → Workouts → Exercises

• Plans contain multiple workouts.
• Workouts contain exercises and can be:
  • MobiFit Workouts: ready-made, created by professionals for the app.
  • My Workouts: User-exclusive, created by them, by AI, or received from someone else (like friends or professionals).
• When opening a plan, the date of the last day on which training sessions were performed in the same week is displayed.

Social and Sharing (Groups + Instagram Stories)

Upon completing a workout or logging an individual activity, a sharing modal appears.

You can publish to the internal group and/or Instagram Stories—with sticker options.

Photos taken vertically are cropped internally to 4:5 for the group feed; on Stories, they go without cropping. The crop marking is displayed at the moment of capture.

Main Stack

• Expo React Native
• Supabase (with Edge Functions)
• Cloudflare R2

About MobiFit

MobiFit is born from a design goal: enabling the first step without friction.

AI acts as a contextual copilot, local-first architecture ensures fluidity in the real world, and the social layer sustains motivation.

All of this in an app with curated original content, multilingual, designed for beginners and experienced trainers alike—in a human, creative, and accessible way.

Next Steps

In development:
wearable integration, such as watches, the Diet page, and new AI agents.
Expanding original media content.