2025
AI Gatekeeping for a Token Launch Platform
Project Info
Hackathon 🏆 Finalist
Submission for Agentic Ethereum 2025
Duration – 10 Days
Links
Website
ETHGlobal Project Page
Roles
Product Management
UX Design
UI Design
Brand Design
Teammates
BE Engineer – Adrian Kögl
FE Engineer – Shouvik Gosh
ML Engineer – Ankit Choudhary
BouncerAI revolutionizes token launches by using AI to evaluate user alignment with project ethos, filtering out bots and speculators while prioritizing genuine supporters. This novel launchpad combines voice-based verification, multi-agent AI assessment, and fair token economics to ensure only those who truly understand and resonate with a project can participate in its growth.
Brand
How It’s Made
TECH STACK
Design prototyped in Figma. Frontend built with Next.js on Vercel. Backend implemented with Node.js (TypeScript, Express, Redis) on a Droplet. Smart contracts developed with Foundry and deployed on Arbitrum. Database managed via Supabase. AI agent framework using LangChain hosted on Autonome.
FRONTEND
Built with Next.js and designed in Figma with TailwindCSS for responsive UI. Users authenticate via Privy with JWT tokens. TradingView Charting Library integrated for real-time token price data visualization. Bouncer avatar designed in Spline to enhance user interaction, with potential future dynamic responses.
BACKEND
TypeScript-based Node.js and Express backend with Redis for session management and Supabase for data storage. Sessions created in Redis after Privy cookie verification. User interactions with project context forwarded to AI agent component and stored for record-keeping. Successful bouncer interactions generate secure signatures with nonces for token allocation. A continuous script monitors projects and updates prices for real-time chart rendering.
DATABASE
Supabase stores user metadata, project data, bouncer configurations, project interaction data, and historical price data for every project contract.
SMART CONTRACTS
Factory contract deploys unique token contracts per project using linear bonding curves. Each contract holds both ETH and corresponding tokens until maturity. Transfer restrictions prevent secondary market bypassing. Core functions include buy() (signature-validated purchases on bonding curve), sell() (allowing users to sell back tokens), and deployLiquidity() (deploying ETH plus tokens to Uniswap once the curve matures).
AI AGENT FRAMEWORK
LangChain-based AI bouncer structured into two workflows:
Interaction Workflow: Multiple parallel evaluation agents (Knowledge Agent and Vibe Agent) calculate scores and draft follow-up questions. A Question Agent then refines the most relevant question based on project style. The adaptive questioning system typically asks 3-5 questions, ending early for weak responses or strategically selecting between Knowledge or Vibe questions.
Briefing Workflow: Users configure the bouncer by providing text summaries/whitepapers, descriptive adjectives for vibe checks, and style adjectives for Question Agent’s tone. System prompts are engineered to align the bouncer with project brand, style, and difficulty. A future Briefing Agent will interact directly with users during project creation.