Top Builders

Explore the top contributors showcasing the highest number of app submissions within our community.

OpenAI Codex

OpenAI Codex is an artificial intelligence model developed by OpenAI. It parses natural language and generates code in response. It is used to power GitHub Copilot, a programming autocompletion tool. Codex is a descendant of OpenAI's GPT-3 model, fine-tuned for use in programming applications. OpenAI has released an API for Codex in closed beta. Based on GPT-3, a neural network trained on text, Codex has additionally been trained on 159 gigabytes of Python code from 54 million GitHub repositories. You can find more information here https://openai.com/blog/openai-codex/

General
Relese dateAugust 31, 2021
AuthorOpenAI
Repository-
TypeAutoregressive, Transformer, Language model

Start building with Codex

We have collected the best Codex libraries and resources to help you get started to build with Codex today. To see what others are building with Codex, check out the community built Codex Use Cases and Applications.

Boilerplates

Kickstart your development with a Codex based boilerplate. Boilerplates is a great way to headstart when building your next project with Codex.

  • Codex Boilerplate Create a function just by typing what it should do, with help of OpenAI Codex.

Libraries

A curated list of libraries and technologies to help you build great projects with Codex.

Edit on GitHub

OpenAI Codex AI technology Hackathon projects

Discover innovative solutions crafted with OpenAI Codex AI technology, developed by our community members during our engaging hackathons.

Vartovii Sentinel-8004

Vartovii Sentinel-8004

Vartovii Sentinel-8004 is a signed control layer for autonomous trading agents. It is not another trading bot. Instead, Sentinel evaluates each trade intent before execution and returns an auditable ALLOW, DENY, or ALLOW_WITH_DOWNSIZE verdict that can be inspected by judges and operators. The public submission focuses on a narrow, production-shaped MVP for ERC-8004-style trade authorization. The hosted demo shows the end-to-end proof flow: a Trade Intent is submitted, Sentinel evaluates it under deterministic policy rules, produces a signed verdict, attaches a validation artifact, and exposes a permit verification result. The flow is fail-closed by design, so missing or invalid conditions do not silently pass. The app is organized for judge clarity. The root URL is a submission hub, and the live walkthrough is available on a stable /judge path. Judges can inspect canonical scenarios, review the intent payload, verdict, validation artifact, and permit status, and compare allow, deny, and constrained execution outcomes. The goal of this MVP is to show that autonomous trading agents can be gated by signed, reviewable trade-permit decisions before capital moves.

Ghetto-Bird Ai Pet

Ghetto-Bird Ai Pet

Ghetto-Bird AI Pet is an AI-powered virtual companion inspired by the creativity, resilience, and culture of urban communities. The project reimagines the digital pet experience by combining conversational AI, personality modeling, and adaptive learning to create a dynamic companion that evolves with its user. Unlike traditional virtual pets, Ghetto-Bird AI Pet is emotionally responsive and culturally aware. It learns from interactions, develops preferences, remembers conversations, and adapts its tone and personality over time. The AI pet can offer motivation, storytelling, humor, life advice, and entertainment, creating a personalized experience for each user. The project aims to make AI companionship more relatable and expressive by blending culture, technology, and emotional intelligence. Whether users are looking for a friend, a hype partner, or a daily check-in companion, Ghetto-Bird AI Pet grows alongside them — becoming smarter, more intuitive, and more uniquely theirs over time.

TradeWise AI

TradeWise AI

Traders face information overload and emotional bias when making market decisions. TradeWise AI addresses this by combining automated technical market intelligence with behavioral insights and social content generation. The platform computes indicators (SMA, RSI, ATR, momentum), identifies patterns, and produces clear, human-readable explanations. It analyzes trader behavior—win rate, holding time, trade frequency—and provides psychology-driven recommendations and trade recovery probabilities. TradeWise AI also includes a RAG-backed search, on-demand content creation for social platforms, real-time alerts, and demo/offline modes for safe exploration. Built as an MVP, it’s aimed at retail traders, trading coaches, and small teams who want actionable analytics, better discipline, and shareable market narratives—while emphasizing education and sound risk management rather than guaranteed predictions.

KYC Agent

KYC Agent

KYC Agent is an intelligent, end-to-end Know-Your-Customer compliance platform built for Deriv. It replaces slow, manual document verification with AI-driven automation while keeping humans in the loop for high-risk cases. How it works: Users upload country-specific identity documents (CNIC, Emirates ID, Passport, Aadhaar/PAN) and proof-of-address through a guided multi-step onboarding wizard. Google Gemini 2.5 Flash performs real-time OCR extraction and document quality analysis — detecting 20+ issues like blur, glare, cropped corners, and potential manipulation. Extracted data is cross-validated against user-submitted form fields to catch mismatches and inconsistencies. Hybrid Risk Scoring Engine: A two-tier risk assessment combines deterministic rule-based checks (data mismatches, expired documents, missing fields, name inconsistencies with transliteration awareness) with Gemini-powered AI analysis for nuanced fraud detection. Submissions are auto-routed: LOW risk → auto-approved, MEDIUM → manual review queue, HIGH → flagged for expert review. Compliance Dashboard: An internal admin tool gives compliance officers a real-time queue of submissions with risk-sorted views, side-by-side form-vs-OCR comparison, detailed risk factor explanations, and one-click approve/reject actions with audit trail notes. Key differentiators: Country-aware intelligence with side-specific OCR (e.g., CNIC back = address extraction) Explainable risk scores — every flag comes with reasoning and severity Graceful Deriv WebSocket API integration with demo-mode fallback Production-grade architecture (FastAPI + Streamlit) with zero external infrastructure dependencies Built with: Streamlit, FastAPI, Google Gemini 2.5 Flash, Deriv WebSocket API, Pydantic, Python

CAREN - AI-Powered Financial Crime Detection

CAREN - AI-Powered Financial Crime Detection

CAREN (Credit Analysis & Risk Evaluation Network) is a next-generation AI-powered transaction monitoring system designed to detect financial crimes in real-time while dramatically reducing false positives that overwhelm compliance teams. The system employs an ensemble of five machine learning models—XGBoost, Random Forest, Logistic Regression, K-Nearest Neighbors, and AdaBoost Decision Tree—trained on anonymized PCA-transformed transaction features (V1-V28). This architecture achieves 99.94% accuracy, 94.12% precision, and 81.63% recall, with an AUC-ROC of 98.21%. Key capabilities include: • Real-time fraud scoring with sub-50ms latency per transaction • Intelligent alert prioritization that turns thousands of weekly alerts into high-confidence cases • Multi-dimensional risk analysis combining velocity patterns, geographic anomalies, and amount deviations • Visual investigation dashboard with transaction timelines and evidence summaries • Configurable detection thresholds for different risk appetites The platform features a comprehensive dashboard for fraud analysts with live transaction monitoring, severity-based alert management, ML model performance analytics, and an AI-powered fraud analyzer that explains predictions using feature importance visualization. Built for scale, CAREN processes transactions in real-time and provides actionable intelligence that reduces investigation time from hours to minutes while maintaining the highest detection accuracy.

Deriv AI Shield

Deriv AI Shield

As organizations rapidly deploy AI agents and LLM-powered applications, new security risks emerge that traditional web security tools cannot address. Prompt injection and adversarial inputs can manipulate AI systems into leaking sensitive data, bypassing safeguards, or executing unintended actions. Since LLMs are designed to interpret natural language flexibly, distinguishing malicious intent from legitimate user input becomes a major challenge. Deriv AI Shield addresses this gap by introducing an AI-native Web Application Firewall tailored specifically for AI systems. The platform monitors and filters user prompts before they reach the LLM, detects jailbreak and prompt injection attempts, prevents sensitive data leakage through output guarding, and continuously analyzes behavior patterns to detect anomalies. In addition, the system includes an autonomous pentesting module that simulates real attack scenarios to identify vulnerabilities and measure security resilience. It generates actionable security insights and mitigation recommendations, enabling teams to strengthen defenses proactively. By combining real-time guardrails, behavioral monitoring, and continuous AI-driven security testing, Deriv AI Shield enables organizations to deploy AI agents safely in production environments without compromising functionality or user experience.