AI agents debate your idea — then build the landing page.
Grand Prize Winner — Build with TRAE Seoul Hackathon 2026
Agent Arena is a multi-agent debate platform where AI agents autonomously discuss, challenge, and refine your business idea — then generate a landing page from the result.
You throw in a topic. A PM agent (always present) and domain-expert agents (auto-generated per topic) debate in a Slack-style chat UI. Agents retire and new specialists spawn as the discussion evolves. When the debate concludes, the chat view transforms into a fully generated landing page.
| Feature | Description |
|---|---|
| Slack-style Debate UI | Real-time multi-agent discussion with typing indicators, join/leave events |
| Auto Agent Spawning | Domain experts are generated based on the topic (marketer, tech lead, UX researcher, etc.) |
| Agent Retirement & Replacement | After each round, the orchestrator retires agents and spawns new specialists |
| PM as Anchor | The PM agent keeps the debate on-topic and user-focused |
| Landing Page Generation | Debate conclusion triggers full HTML landing page generation |
| Page Swap Transition | Animated transition from chat to the generated landing page |
| Debate History | Browse and revisit past debates and their generated pages |
| Preset Topics | Quick-start with curated topic suggestions |
Agent Arena demonstrates 7+ agent design patterns:
| Pattern | Usage |
|---|---|
| Orchestrator | Controls debate flow, round management, agent lifecycle |
| Specialist Agents | Domain experts auto-generated per topic |
| Persistent Agent (PM) | Always-on anchor that prevents scope creep |
| Agent Spawning | New experts created mid-debate when the topic shifts |
| Agent Retirement | Graceful exit with handoff message when no longer relevant |
| Multi-turn Debate | Structured rounds with analysis between turns |
| Result Synthesis | Final round produces structured output for page generation |
| Category | Technology |
|---|---|
| Framework | Next.js 16 (App Router) |
| Language | TypeScript (strict) |
| AI | OpenAI GPT-4o |
| UI | Tailwind CSS v4, Framer Motion |
| State | React hooks |
| Export | Archiver (ZIP download) |
git clone https://github.com/wigtn/wigent.git
cd wigent
npm install
npm run devSet your OpenAI API key in the environment:
export OPENAI_API_KEY=sk-...Open http://localhost:3000, enter a topic or pick a preset, and watch the agents debate.
wigent/
├── src/
│ ├── app/
│ │ ├── api/
│ │ │ ├── debate/ # Debate orchestration API
│ │ │ └── export/ # Landing page export (ZIP)
│ │ ├── page.tsx # Main page (topic input → debate → landing)
│ │ └── layout.tsx # Root layout
│ ├── components/
│ │ ├── chat/ # Slack-style chat UI components
│ │ ├── TopicInput.tsx # Topic input form
│ │ ├── PresetButtons.tsx # Preset topic suggestions
│ │ ├── LandingPageView.tsx # Generated landing page renderer
│ │ ├── HistoryList.tsx # Debate history browser
│ │ └── PageTransition.tsx # Chat → Landing page transition
│ ├── hooks/ # useDebate, useHistory
│ └── lib/ # Types, utilities
├── docs/ # Documentation
├── PRD.md # Product Requirements Document
└── package.json
1. User enters topic (or picks preset)
2. Orchestrator creates PM agent + topic-specific expert
3. Agents debate in rounds (Slack-style UI)
4. After each round: orchestrator analyzes → may retire/spawn agents
5. Final round: structured result synthesis
6. GPT-4o generates full landing page HTML
7. Animated transition: chat → landing page
8. Auto-saved to history
MIT
This project is built and maintained by WIGTN Crew —
an AI-native open-source research crew based in Korea.
We build practical, domain-specialized AI tools. Fast prototyping, strong engineering, shipping real things.
| 🌐 Website | wigtn.com |
| 🐙 GitHub | github.com/wigtn |
| 🤗 HuggingFace | huggingface.co/Wigtn |
| 📦 NPM | npmjs.com/org/wigtn |
| Project | Description | Status |
|---|---|---|
| WigtnOCR | VLM-based Korean government document parser | Research |
| WIGVO | Real-time PSTN voice translation (Korean↔English) | Research |
| Claude Code Plugin | Multi-agent parallel execution ecosystem | Open Source |
Results speak louder than resumes.