BigBowl | Devpost

Big Bowl

Inspiration 💡

BigBowl was created to democratize smart contract security and bring AI agents onto the blockchain. The core idea is to let anyone audit contracts instantly and own their custom AI auditor as a transferable NFT. This addresses the need for accessible, transparent, and user-owned tools in the Web3 space.

What it Does ⚙️

BigBowl is a browser-based platform that performs security audits on Ethereum smart contracts and allows users to mint custom AI agents as NFTs. It analyzes code, generates a Safety Index and threat summary, and enables real-time interaction with the user's on-chain AI agent. The end-to-end process involves connecting a wallet, analyzing a contract, and signing a transaction to own an AI agent.

How We Built It 🛠️

The platform is built with a FastAPI backend for core business logic and analysis, powered by a custom Spoon AI agent framework for intelligence and orchestration. The frontend is a React/TypeScript application that connects to the backend via HTTP/WebSockets and integrates with MetaMask using Web3.js for on-chain transactions like NFT minting. The architecture uses external services like Sourcify to fetch contract source code for analysis and various LLM providers for generating human-readable summaries.

## Challenges We Ran Into 🚧 Integrating the diverse set of technologies, especially ensuring seamless, secure wallet-to-backend communication for NFT minting, presented a significant challenge. We also faced complexity in designing the hybrid on-chain/off-chain execution for the AI agents, balancing trustless execution with the high resource cost of LLMs. Additionally, structuring the static analysis and pattern detection to be both fast and accurate required careful engineering.

Accomplishments That We're Proud Of ✨

We are most proud of successfully implementing the AI Agent NFT minting workflow, proving that a specialized AI entity can be an ownable, blockchain-native asset. Achieving the full end-to-end user journey—from contract analysis to on-chain agent deployment—was a major technical feat. Finally, developing the Spoon AI agent framework as a modular core for intelligent, customizable workflows is a significant accomplishment.

What We Learned 🧠

We gained deep expertise in integrating AI agent frameworks with Web3 wallets and decentralized transaction signing protocols like MetaMask. The process taught us the value of a modular, microservices-oriented architecture (FastAPI, LLM Service, Static Analyzer) for handling complex, multi-step operations. We also better understood the trade-offs between purely on-chain vs. hybrid agent execution and how to manage the associated gas costs and latency.

What's Next for BigBowl 🚀

The immediate next steps are to expand the suite of specialized AI agents (e.g., enhanced Security Agent, multi-chain Trading Agent) and introduce a marketplace for trading and delegating agent ownership. We plan to integrate more advanced, dynamic analysis tools beyond the current static and LLM analysis to catch a wider range of vulnerabilities. Ultimately, BigBowl aims to become the de-facto standard for user-owned, decentralized AI security tools.