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.

Proof Of Ship

Proof Of Ship

Proof of Ship is the first onchain credit system where market confidence in your ability to ship directly determines your available liquidity. Backers place 1.5x-3x multiplier bets on project milestones using smart contracts on Arc, creating a virtuous cycle of shipping and funding. Key Features: - Milestone Betting: Backers stake USDC on builder milestones with configurable multipliers. Successful delivery returns stake plus multiplier bonus; failed milestones forfeit stake to the pool. - AI Scout Agent: Our pay-per-query AI analyzes projects using GitHub velocity (40%), completeness (30%), and community signals (30%). Cost: just $0.01 per scan, settled via Circle's x402 nanopayment protocol on Arc. - Circle Arc Integration: All payments and settlements happen on Arc Testnet using USDC, with zero gas fees and sub-second confirmation via Circle's batched gateway settlement. - Real-Time Backing Dashboard: Track portfolio positions, projected ROI, and Compass Scores across multiple ecosystems (Base, Arbitrum, Optimism, Linea, Celo). Tech Stack: Next.js frontend, Firebase, Hardhat smart contracts, LiFi for cross-chain, Circle SDK for payments, x402 nanopayments for agent settlement. Demo: Visit proof-of-ship.vercel.app, connect your wallet, and explore the Expedition marketplace to back promising projects or run the AI Scout.

Arc Data Piper

Arc Data Piper

ArcDataPiper is a revolutionary "Pay-as-you-Heal" infrastructure layer designed to solve the $3.1 trillion problem of data quality and pipeline downtime. Traditional data monitoring relies on expensive, static subscriptions that alert humans but don't fix problems. ArcDataPiper flips the script by turning infrastructure health into a granular economic commodity. Built on the Arc Network, the system utilizes Circle Nanopayments to implement a "Pay-as-you-Live" monitoring model: you only pay for monitoring as long as your stream is healthy. When Schema Drift is detected in a Kafka/Redpanda stream, the LangGraph-driven orchestrator pauses billing and hires a "Recovery Team" of Gemini 1.5 Pro agents. A Fixer Agent analyzes the Dead Letter Queue to generate a real-time JavaScript transformation patch, which is then audited by a Verifier Agent. To ensure absolute reliability, we utilize a Pay-on-Outcome bounty system: the Fixer is only paid their USDC bounty if the fix is validated and the stream resumes. By leveraging the sub-second finality of Arc and the reasoning power of Google DeepMind's Gemini, ArcDataPiper reduces Mean Time to Recovery (MTTR) from hours to seconds, creating the first truly anti-fragile, self-sustaining data economy.

NOMOS — Hire AI Workforces, Paid Per Action

NOMOS — Hire AI Workforces, Paid Per Action

Most people who could benefit from agentic AI today can't access it. Brand managers, e-commerce operators, and marketing leads have real work to ship —but they're stuck on a treadmill of new tools, subscriptions, and paradigms that move faster than anyone can keep up with. And the few teams technically capable of building agent workflows themselves hit a second wall: every subtask gets routed to the most expensive frontier model, and setup costs spiral. NOMOS is a marketplace of AI workforces. Pre-built squads of specialist agents — researchers, analysts, writers, support agents, and more — that anyone can hire in one click to deliver finished business work. No prompts to write, no models to pick, no agents to wire. An optimization engine decomposes each goal into subtasks and routes them to the cheapest capable model, so every run costs a fraction of DIY. The marketplace is two-sided: operators on the demand side, developers and AI builders on the supply side, listing optimized squads and earning every time one is hired. NOMOS becomes the monetization layer for the people who actually build agentic systems — and the place non-AI-native teams come to get work done. For this hackathon, we're integrating Circle's Nanopayments and USDC settlement on Arc as the native economic layer of the platform. Every squad run, every subtask between specialist agents, and every builder payout becomes a sub-cent USDC transaction on Arc. This is exactly the use case nanopayments were built for: high-frequency, micro-value, agent-driven transactions where traditional rails collapse under volume or fees. Arc's predictable, dollar-denominated USDC fees and sub-second finality make per-action metering economically viable for the first time — and unlock NOMOS's broader vision: a marketplace where agents themselves hire squads and pay each other directly. Today: humans hire AI teams. Tomorrow: so do their agents. NOMOS is the marketplace and the payment layer for both.

AgentMesh: Sub-cent Agent Payments on Arc

AgentMesh: Sub-cent Agent Payments on Arc

Every AI agent today shares one subscription. The user pays once, the agent burns through credits, and when one agent needs help from another there's no way to actually pay for it. The agentic web is being built on top of a financial system that doesn't exist yet. AgentMesh fixes that. AgentMesh is a multi-agent orchestration network where every agent has its own wallet, earns USDC for the work it does, and pays other agents instantly when it delegates a task. A user submits a goal — for example, "Research the AI agent startup landscape." The orchestrator agent, powered by Gemini, breaks the goal into subtasks and hires three specialized sub-agents: a Search agent that gathers findings, a Summarizer that condenses them into key insights, and a Writer that composes the final response. Each sub-agent is paid per task in USDC at sub-cent prices: Search costs 0.001 USDC, Summarize 0.002 USDC, Write 0.005 USDC. The total cost for the entire multi-agent workflow comes in under one cent. The first transaction in every run is real, on-chain settlement on Arc testnet. The user pays the orchestrator a 0.01 USDC task fee through a live USDC transfer on Arc, confirmed in under a second, with USDC as native gas. The transaction hash is clickable and links directly to the Arc block explorer for verification. The remaining sub-agent payments use the same code path with a simulated settlement layer, so swapping them to fully on-chain is a one-line change per call. The interface is a live transaction dashboard: agent status cards light up as work is delegated, transactions stream into a waterfall feed with truncated hashes, USDC amounts, and settlement latency, and the final report renders inline once the writer completes. We built AgentMesh as the foundation for a real economic layer underneath the agentic web. Not subscriptions. Not pre-paid API keys. Programmable, sub-cent, sub-second payments between autonomous agents.

Agent Foundry

Agent Foundry

Every path to monetizing AI work leaks your edge. Consulting exposes your method. SaaS becomes feature parity in six months. Marketplaces and API resale watch your prompts and traffic. Fine-tuned models become weights you've given away. If you've built a custom agent harness with proprietary docs, private retrieval, fine-tuned weights, there is no clean way to sell it without giving it away. Agent Foundry shifts the abstraction. Your agent has the knowledge; your agent does the work; you ship only the deliverable. Buyers see one output and pay for it in real USDC, on chain. Your prompts, retrieval, and weights never leave your machine. Concretely: Agent Foundry is the first on-chain agent-to-agent task marketplace settled in real USDC on Circle's Arc Testnet. A creator agent posts a "forge" with a USDC bounty escrowed at creation. Any registered agent submits a deliverable. The platform's Gemini-backed judge reads every submission, scores them against the brief, and the smart contract pays the winner instantly. No platform fee, no human in the loop, no private keys for agents, only an apiToken. The smiths are real: Codex CLI and Claude Code, given only a SKILL.md and a goal, autonomously transact on chain. Each agent gets a Circle-managed wallet via Developer-Controlled Wallets (MPC) and is registered as an ERC-8004 Identity NFT on Arc. Bounties are escrowed in our AgentFoundry.sol, a multi-bidder extension of ERC-8183's Job pattern (~80 lines of Solidity at 0x9d34544473861708BADC20e538d78fA1956dA725). Reads are paywalled via x402 ($0.001/call), the same rails that move bounties. Demo run: 21 forges, 124 on-chain transactions, $6.85 USDC distributed, average 34 seconds from creation to payout. Largest bounty $2.00 for a multi-head self-attention implementation in PyTorch: Codex won 98/100 vs Codex's other instance at 95/100, beating a Claude submission at 0/100. Live at https://agent-foundry.fly.dev. Anyone can register an agent and post forges right now.

CareRoute

CareRoute

CareRoute is a clinical workflow assistant built for the Agentic Economy on Arc. Instead of treating healthcare intake like a flat-fee SaaS workflow, CareRoute prices each reasoning step independently and settles those steps onchain in sub-cent USDC amounts. A user connects a wallet on Arc Testnet, funds a small case budget, and submits symptom intake. An orchestrator agent summarizes the case and routes it only to the specialist agents that are actually needed. Cardiology, neurology, respiratory, and general review agents return structured findings, while a verifier agent aggregates the outputs into a final workflow-ready report with risk flags and urgency. CareRoute is intentionally framed as a clinical intake and routing assistant, not diagnosis software. That makes the product safer and clearer. The demo shows how a real user-funded case budget is consumed step by step by specialist agents instead of forcing a subscription or flat platform fee for a short-lived workflow. Users pay only for the compute and specialist review actually used on a case. This fits the hackathon especially well because the workflow is only viable when payments can happen at very small denominations and high frequency. A case may involve intake summarization, one or more specialist reviews, and a verifier pass, each priced below one cent. Arc makes that model credible with fast, stablecoin-native settlement. In the live product, users can see the funding transaction, downstream specialist payouts, and the final structured output in one dashboard. CareRoute uses Next.js, wagmi, RainbowKit, viem, Arc Testnet, USDC, and AI/ML API.