CourtSync

one text. four players. game on.

The AI agent that coordinates pickleball matches through WhatsApp. Text your skill level, preferred format, and when you want to play. CourtSync handles the rest.

The problem

Pickleball is the fastest-growing sport in America. 19.8 million players. 311% growth since 2020. And the coordination is still done in group chats, spreadsheets, and Facebook threads.

Players waste more time finding a match than playing one.

You text 15 people to fill a doubles game
Half don't respond, half are the wrong skill level
By the time you have 4 confirmed, the court slot is gone
You do it again tomorrow

CourtSync fixes this in one message.

You:        "3.5 doubles tomorrow 6pm"
CourtSync:  "Found 3 players at the 3.0-4.0 level. Reaching out now."
CourtSync:  "Match confirmed! 4 players. Tomorrow 6pm. See you on the court."

No app to download. No account to create. Just WhatsApp -- the app 60% of pickleball players already use daily.

How it works

┌────────────────────────────────────────────────────────────────┐
│                                                                │
│   Player sends WhatsApp message                                │
│   "3.5 doubles tomorrow 6pm"                                   │
│                                                                │
│   ┌──────────┐    ┌───────────────┐    ┌────────────────────┐  │
│   │  Twilio   │───▶│  Cloud Run    │───▶│  Pub/Sub Queue     │  │
│   │  Webhook  │    │  API Gateway  │    │  (async routing)   │  │
│   └──────────┘    └───────────────┘    └────────┬───────────┘  │
│                                                  │              │
│                   ┌──────────────────────────────┘              │
│                   ▼                                             │
│   ┌───────────────────────────────────────────────────────┐    │
│   │              Message Handler                          │    │
│   │                                                       │    │
│   │  ┌─────────────────┐    ┌──────────────────────────┐  │    │
│   │  │  Gemini 2.0     │    │  Skill Matcher           │  │    │
│   │  │  Flash          │    │                          │  │    │
│   │  │                 │    │  Composite score:        │  │    │
│   │  │  Parse intent   │    │  - Skill proximity (40%) │  │    │
│   │  │  Extract skill  │    │  - Distance (30%)        │  │    │
│   │  │  Extract time   │    │  - Reliability (20%)     │  │    │
│   │  │  Extract format │    │  - Engagement (10%)      │  │    │
│   │  └────────┬────────┘    └────────────┬─────────────┘  │    │
│   │           │                          │                │    │
│   │           └──────────┬───────────────┘                │    │
│   │                      ▼                                │    │
│   │           ┌──────────────────────┐                    │    │
│   │           │     Firestore        │                    │    │
│   │           │  players | matches   │                    │    │
│   │           └──────────────────────┘                    │    │
│   └───────────────────────────────────────────────────────┘    │
│                                                                │
│   ┌──────────────────┐    ┌──────────────────────────────┐    │
│   │  Notification     │    │  Reminder Job               │    │
│   │  Sender           │    │  (hourly via Cloud          │    │
│   │  (Pub/Sub → SMS)  │    │   Scheduler)                │    │
│   └──────────────────┘    └──────────────────────────────┘    │
│                                                                │
└────────────────────────────────────────────────────────────────┘

The matching engine

When you say "3.5 doubles tomorrow 6pm", CourtSync:

Parses your message with Gemini 2.0 Flash -- extracts skill level, format, time preference, location
Queries all onboarded players within your DUPR range (default +/-0.5)
Scores each candidate on a 0-100 composite:

Factor	Weight	How it works
Skill proximity	40%	Closer DUPR rating = higher score
Geographic distance	30%	Within 25km preferred
Reliability	20%	Based on no-show history
Engagement	10%	Recent match activity

Invites the top 5 candidates via WhatsApp
Tracks YES/NO responses, moves players through the pipeline
Confirms once the group is full and notifies everyone

The entire flow happens over WhatsApp. No app. No login. No friction.

The streak rules

CourtSync tracks reliability because nobody wants to organize a foursome and have someone ghost:

Reliability score starts at 1.0 (100%)
No-shows decrease your score -- you get matched less often
Consistent players surface first in every search
The system rewards people who show up

Tech stack

Layer	Technology	Why
Messaging	Twilio WhatsApp API	Instant sandbox access, production-ready
AI	Gemini 2.0 Flash (Vertex AI)	Fast intent parsing, JSON output, low cost
API Gateway	Cloud Run (FastAPI)	Auto-scaling, 10s timeout, signature verification
Workers	Cloud Functions (Gen2)	Event-driven, Pub/Sub triggered
Database	Firestore	NoSQL, real-time, zero ops
Queue	Google Cloud Pub/Sub	Async message routing between services
Scheduler	Cloud Scheduler	Hourly reminder jobs
Geocoding	geopy	Distance calculations for proximity matching

Cost at beta (100 users, 400 matches/month): ~$43/month.

Security

CourtSync handles phone numbers, locations, and message content. Security is not optional.

What we enforce

Protection	Implementation
Webhook authentication	Twilio signature verification (HMAC) -- always on in production
Phone validation	Regex `^\+?1?\d{10,15}$` on every input, rejects malformed numbers
Rate limiting	20 messages/minute per phone number
Prompt injection defense	User input sanitized before AI prompt, phone numbers stripped from AI context
Message limits	2000 chars inbound, 1600 chars outbound
Log redaction	Phone numbers masked in all logs (`+121****`)
Secret management	All credentials in GCP Secret Manager, never in code
API docs hidden	`/docs` and `/openapi.json` disabled in production
Input sanitization	Profile names truncated, message bodies length-limited

What we don't store

No passwords (WhatsApp identity only)
No payment data (not yet)
No message content sent to third parties (Gemini runs on Vertex AI within GCP)

See SECURITY.md for our vulnerability disclosure policy.

Quick start

Prerequisites

Google Cloud account with billing enabled
Twilio account with WhatsApp Sandbox
Python 3.11+
gcloud CLI authenticated

1. Clone and configure

git clone https://github.com/odominguez7/Courtsync-Pickleboard.git
cd Courtsync-Pickleboard
cp .env.yaml.example .env.yaml   # fill in your credentials

2. Bootstrap GCP

export PROJECT_ID="your-project-id"
export BILLING_ACCOUNT="XXXXXX-XXXXXX-XXXXXX"
./scripts/setup_gcp.sh

This creates the project, enables APIs, sets up Firestore, Pub/Sub topics, and Secret Manager entries.

3. Deploy

# Lean MVP (single Cloud Function — ship in minutes)
./scripts/deploy.sh lean

# Production (Cloud Run + Cloud Functions + Scheduler)
./scripts/deploy.sh production

4. Connect Twilio

In Twilio Console > Messaging > WhatsApp Sandbox, set your webhook URL to the deployed function URL.

5. Test

Open WhatsApp, message your sandbox number:

3.5 doubles tomorrow 6pm

Project structure

courtsync/
├── function/                         Lean MVP (single Cloud Function)
│   ├── main.py                       Webhook entry point
│   ├── coordinator.py                AI coordination engine (Gemini + Firestore)
│   ├── matcher.py                    Skill + location + reliability matching
│   └── requirements.txt
│
├── config/
│   └── prompts.py                    Gemini system prompts + few-shot examples
│
├── infrastructure/                   Production architecture
│   ├── api/                          Cloud Run API Gateway (FastAPI)
│   │   ├── main.py                   Webhook + signature verification + rate limiting
│   │   ├── Dockerfile
│   │   └── requirements.txt
│   ├── functions/
│   │   ├── message_handler/          Pub/Sub → Coordinator → Notifications
│   │   ├── negotiation_engine/       AI schedule negotiation (future)
│   │   └── notification_sender/      Pub/Sub → Twilio WhatsApp
│   └── jobs/
│       └── send_reminders/           Hourly match reminder cron
│
├── scripts/
│   ├── setup_gcp.sh                  One-time GCP project bootstrap
│   ├── deploy.sh                     Deploy lean or production
│   └── test_local.sh                 Local development
│
├── SECURITY.md                       Vulnerability disclosure policy
├── .env.yaml.example                 Environment template
└── .gitignore

Database schema

`players` collection

{
  "phone": "+12125551234",
  "profile": {
    "name": "Sam Chen",
    "dupr_rating": 3.5,
    "location": { "lat": 42.36, "lng": -71.05, "city": "Boston" }
  },
  "preferences": {
    "formats": ["doubles", "mixed_doubles"],
    "max_drive_minutes": 15
  },
  "stats": {
    "matches_played": 12,
    "reliability_score": 0.95
  },
  "active_match_id": null,
  "onboarding_complete": true
}

`matches` collection

{
  "match_id": "auto-generated",
  "status": "seeking_players | confirmed | completed | cancelled",
  "format": "doubles",
  "skill_range": { "min": 3.0, "max": 4.0, "target": 3.5 },
  "players": {
    "needed": 4,
    "confirmed": ["+1...", "+1..."],
    "pending": ["+1..."],
    "declined": []
  },
  "schedule": {
    "time_preference": "tomorrow 6pm",
    "duration_minutes": 90
  }
}

The market

Stat	Number	Source
US pickleball players	19.8 million	SFIA 2024
Growth since 2020	311%	SFIA 2024
Players aged 55+	60%+	USA Pickleball
Public courts	10,300+	Places2Play
USAPA ambassadors	2,200 (400+ player lists each)	USA Pickleball
YoY growth rate	44%	APP Tour

Primary persona: "Suburban Social Striver Sam"

30-45, full-time job, limited free time
Uses WhatsApp daily
Doesn't want another app
Wants to maximize court time per hour of coordination effort

Target markets (priority): Florida, California, Texas, Arizona, Utah

Research foundation

Built on peer-reviewed research, not guesswork:

Heo & Ryu (2023): Social connection is the primary retention driver in pickleball
Lozano et al. (2025): Match activity profiles enable scheduling optimization
Prieto-Lage et al. (2024): Performance analytics by court zone improve skill matching
WHO-5 Wellbeing Index: 3x/week play shows measurable wellbeing improvement
SFIA 2024: Market sizing and demographic data

See md files MIT Libraries (R&D)/ for full research synthesis.

Revenue model

Tier	Price	Matches/month	Features
Free	$0	4	Basic matching, WhatsApp coordination
Premium	$9.99/mo	Unlimited	Priority matching, analytics, calendar sync
Ambassador	$49/mo	Unlimited	League tools, bulk import, dashboard, branding

Scale	Users	Monthly cost	Projected ARR
Beta	100	~$43	--
Growth	10K	~$4,150	$50K-$100K
Scale	100K	~$2,000*	$500K-$1M

*Meta Cloud API replaces Twilio per-message costs at scale.

Roadmap

Phase 1 -- Beta (Weeks 1-4)

WhatsApp webhook + Twilio integration
Gemini AI intent parsing
Skill-based player matching (DUPR +/-0.5)
Match creation + player notification
YES/NO response handling
Match confirmation flow
Security hardening (auth, rate limiting, input validation)
Google Calendar invite integration
First 100 users via USAPA ambassador network

Phase 2 -- Growth (Months 2-3)

Court database with geo search
Negotiation engine (flexible scheduling)
Reliability scoring (no-show tracking)
Premium tier subscription
PostHog analytics dashboard
1,000 users

Phase 3 -- Scale (Months 4-6)

The big objective

Pickleball doesn't need another app. It needs infrastructure.

19.8 million people play this sport and the coordination layer is still group chats and spreadsheets. That's not a feature gap -- it's a missing layer of the stack.

CourtSync is that layer. WhatsApp-native because that's where the players already are. AI-powered because matching by skill, location, reliability, and schedule is a problem that gets better with data, not worse. Open source because the best infrastructure is built in the open.

The goal is simple: make it easier to play than to not play. If finding a match takes one text message instead of fifteen, more people play. More people play, more people stay. More people stay, the sport grows.

We're not building a social network. We're not building a fitness app. We're building the coordination protocol for the fastest-growing sport in America.

One text. Four players. Game on.

Built by Omar -- MIT Sloan '26

Coordination so good, you just have to show up and play.

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Repository files navigation

CourtSync

The problem

How it works

The matching engine

The streak rules

Tech stack

Security

What we enforce

What we don't store

Quick start

Prerequisites

1. Clone and configure

2. Bootstrap GCP

3. Deploy

4. Connect Twilio

5. Test

Project structure

Database schema

`players` collection

`matches` collection

The market

Research foundation

Revenue model

Roadmap

Phase 1 -- Beta (Weeks 1-4)

Phase 2 -- Growth (Months 2-3)

Phase 3 -- Scale (Months 4-6)

The big objective

About

Uh oh!

Releases

Packages

Uh oh!

Contributors

Uh oh!

Languages

Name		Name	Last commit message	Last commit date
Latest commit History 5 Commits
Files ready to ship /Courtsync V1		Files ready to ship /Courtsync V1
Product architectures for cursor		Product architectures for cursor
config		config
function		function
infrastructure		infrastructure
md files MIT Libraries (R&D)		md files MIT Libraries (R&D)
scripts		scripts
.env.yaml.example		.env.yaml.example
.gitignore		.gitignore
COURTSYNC_INFRASTRUCTURE.md		COURTSYNC_INFRASTRUCTURE.md
LICENSE		LICENSE
README.md		README.md
SECURITY.md		SECURITY.md

Folders and files

Latest commit

History

Repository files navigation

CourtSync

The problem

How it works

The matching engine

The streak rules

Tech stack

Security

What we enforce

What we don't store

Quick start

Prerequisites

1. Clone and configure

2. Bootstrap GCP

3. Deploy

4. Connect Twilio

5. Test

Project structure

Database schema

players collection

matches collection

The market

Research foundation

Revenue model

Roadmap

Phase 1 -- Beta (Weeks 1-4)

Phase 2 -- Growth (Months 2-3)

Phase 3 -- Scale (Months 4-6)

The big objective

About

Topics

Resources

License

Security policy

Uh oh!

Stars

Watchers

Forks

Releases

Packages 0

Uh oh!

Contributors

Uh oh!

Languages

`players` collection

`matches` collection

Packages