Agent Forge AI Hackathon ハッカソン東京

The era of simple chatbots is over. Real AI applications need to access live information, reason across connected data, execute code, use specialized models, and run reliably in production. The Agent Forge AI Hackathon brings builders together to create and launch working AI agents in one fast-paced afternoon. Using technologies from top AI companies, teams will take an idea from concept to a live demonstration. Build, test, and deploy something real—not just a slide deck or a generic chatbot. The Challenge Build an AI agent—or a coordinated team of agents—that solves a meaningful real-world problem. Your project might automate a complex workflow, conduct live research, reason over connected knowledge, generate or execute code, work across multiple modalities, or create an entirely new agentic experience. Your Agent Forge project should be able to gather information, reason or act autonomously, and demonstrate a clear practical use case. You are free to explore any industry or problem space. The core requirement: build with one or more of the partner technologies below and have a working project ready to demonstrate by the submission deadline. AI Stack Partners Participants will receive access to tools, technical guidance, and resources from our innovation partners: CodeAF by AgentField AI: the open coding harness for open models. Built for DeepSeek, Qwen, GLM and Kimi, and works with the closed frontier models you already pay for, any provider you choose, or a local model. One session, many models: the right model for each call. Your sessions are plain files on your disk. On DeepSWE it fixed the most issues of ten coding harnesses on the same model, at the lowest cost per fix. Apache 2.0. To build autonomous software of your own, AgentField is the open-source infrastructure for it: write agents as Python, TypeScript or Go functions, and its control plane handles orchestration, memory and observability. Daytona: an open source development environment manager and runtime infrastructure designed to streamline software engineering and autonomous agent workflows. By automating workspace creation, Daytona allows developers to spin up standardized, secure environments instantly across any infrastructure, including local machines, remote servers, and major cloud providers. The platform eliminates traditional configuration friction by leveraging containerized setups alongside seamless integrations with popular editors like Visual Studio Code and JetBrains. Furthermore, Daytona provides isolated sandboxes tailored for executing AI generated code safely. By standardizing development environments and securing automated execution, Daytona empowers modern engineering teams to build software rapidly without environmental bottlenecks. Tavily AI: an intelligent search engine designed specifically for large language models, autonomous agents, and retrieval augmented generation workflows. Unlike traditional search platforms built for human browsing, Tavily focuses on delivering clean, highly relevant, and real time factual information directly to artificial intelligence systems without advertisement clutter or unnecessary website markup. The service extracts concise knowledge to minimize token usage, prevent model hallucination, and accelerate multi step research tasks. With seamless integrations across popular developer frameworks like LangChain and LlamaIndex, Tavily serves as the essential real world web connection that helps autonomous software reason accurately and make grounded, reliable decisions. Nosana: a decentralized compute network built on the Solana blockchain that connects artificial intelligence developers with distributed, idle GPU hardware around the globe. By creating an open marketplace for computing power, Nosana provides an affordable and accessible alternative to centralized cloud monopolies like Amazon Web Services and Google Cloud. Independent hardware owners monetize unused consumer and enterprise graphics cards, while builders access cost efficient compute capacity for running model inference and machine learning workloads. Powered by the native NOS token for secure transactions, Nosana democratizes access to scarce hardware and accelerates the deployment of scalable artificial intelligence systems worldwide. Specific participant credits, technical resources, and partner prizes will be announced separately. Team Format Participate individually or form a team of up to six people. We encourage teams to combine different perspectives and skills—including development, design, product, business, and domain expertise—to create more ambitious and complete projects. Don’t have a team yet? Join us anyway. There will be an opportunity to meet other participants and form teams at the event. Judging Criteria Innovation: Does the project introduce a creative or original agentic experience? Real-World Impact: Does it address a genuine problem or meaningful market need? Technical Execution: Is there a functional, live demonstration rather than only a concept? Partner Technology Usage: How effectively and meaningfully are the sponsor technologies integrated? Projects using multiple partner technologies may receive additional consideration. Code-level integrations will be reviewed. Prizes Winning teams will receive exclusive hackathon prizes and partner benefits to help them continue developing and scaling their projects. Agenda 2:00 PM — Check-in & Networking 2:20 PM — Welcome and Hackathon Introduction 2:30 PM — Partner Tech Talks & Building Guidelines 3:10 PM — Team Formation and Project Setup 3:30 PM — Building Session 5:30 PM — Submission Deadline 5:40 PM — Live Demos 6:30 PM — Judging and Networking 6:45 PM — Winners Announced 7:00 PM — Event Close Agenda timings are subject to minor changes. Who Should Come? This event is for people who want to build AI agents that interact with real tools, live information, connected knowledge, models, and compute—not just static datasets. It is for buil