SuiGPT is an AI-powered Move smart contract decompiler built specifically for the Sui blockchain. Its core product, the Move AI Decompiler (MAD), takes compiled smart contract bytecode from non-open-source Sui contracts and uses large language models with prompt engineering to reconstruct logically correct, human-readable, and re-compilable Move source code. It was introduced in a research paper published at the ACM Web Conference 2025 (DOI: 10.1145/3696410.3714790). The tool addresses a real transparency gap in the Sui ecosystem: many deployed smart contracts are not open-source, making independent security auditing impossible with traditional disassemblers. SuiGPT MAD chains the outputs of the Revela decompiler and Move Disassembler through an LLM pipeline, using domain-specific knowledge injection, structured instructions, and 17-example few-shot learning to reduce hallucinations and produce idiomatic Move code. In benchmarks reported by the authors, MAD achieves a 73% recompilation success rate on real-world Sui contracts. A user study with 12 blockchain developers found that MAD significantly reduced auditing workload compared to raw disassembler output, and participants rated its output quality as comparable to the original source code. The frontend and backend are built with Next.js deployed on Vercel using edge functions and serverless architecture for parallel LLM processing. An integrated chatbot assists users in identifying and analyzing specific vulnerabilities in the decompiled code. The tool is accessible via the web at suigpt.tools and is oriented toward smart contract auditors, security researchers, and Sui ecosystem developers who need to inspect contracts they did not write. Key features: - Decompiles Sui Move bytecode into human-readable and re-compilable Move source code - 73% recompilation success rate on real-world non-open-source Sui contracts - LLM prompt engineering pipeline combining Revela decompiler output and Move Disassembler - Integrated chatbot for vulnerability identification and analysis of decompiled contracts - Export of decompiled Move source code for offline analysis - Few-shot learning with 17 curated examples to reduce LLM hallucinations - Serverless Next.js/Vercel architecture enabling parallel contract processing - Peer-reviewed research backing published at ACM Web Conference 2025
Pricing Information for SuiGPT The pricing for the AI agent "SuiGPT" is not available.