ChatSafe
AI-Powered Safety for Onchain Chats.
About
The problem it solves 🛡️ **The Problem It Solves** Group messaging onchain is powerful—but dangerously unmoderated. As blockchain-based communication tools like XMTP grow, so does the risk of abuse: spam, hate speech, scams, and misinformation. Most crypto-native group chats today have no automated moderation, no community safety tooling, and no accountability baked into their infrastructure. This leads to: - **Unsafe environments** for users - **Spam and scam exposure** - **Reputation loss** for communities and protocols - **Manual moderation** burden on admins (if any) ✅ **What ChatSafe Does** ChatSafe is an onchain, AI-powered safety agent for group chats built on XMTP. It brings real-time moderation, accountability, and transparency to decentralized messaging. 🧩 ***Use Cases-*** 🔐 **DAOs & Protocols** — Keep governance and contributor chats clean, safe, and civil 🌐 **Onchain Communities** — Moderate group conversations at scale, automatically 💬 **Web3 Social Platforms** — Offer built-in moderation for user-generated content 🛠️ **Infra Builders **— Add ChatSafe as a plug-in moderation agent for new chat dApps 💡 **How It Makes Life Easier** Without ChatSafe With ChatSafe ✨ 1. - 🚫 Manual moderation ✅ Automated, AI-powered filtering 2. - ❌ No message records ✅ All infractions logged onchain 3. - 🕳️ Zero transparency ✅ Publicly auditable moderation 4. - 😵💫 No trust signals ✅ Onchain user reputation system [https://github.com/0x-Parzival/ChatSafe](url)
How it's made
Challenges we ran into **Challenges I Ran Into** One of the biggest challenges was integrating all the moving parts: XMTP messaging, OpenAI moderation, smart contracts on Base, and a real-time dashboard — all in a full-stack system that had to communicate seamlessly. Specific Hurdles: - **XMTP Listener + Agent Reliability** Building an always-on moderation agent that listens to XMTP group messages was tricky. We had to ensure stable message polling, avoid duplicate processing, and handle intermittent failures gracefully. - **Smart Contract ABI Sync** Keeping the frontend and backend in sync with the smart contract ABI was error-prone. Small mismatches in function signatures caused subtle bugs. We solved this by creating a helper script to copy the compiled ABI to both the agent and dashboard automatically. - **Rate-Limiting + Token Usage (OpenAI)** The moderation AI (OpenAI's model) needs to be called responsibly. We had to implement logic to only call the model when messages actually needed filtering — and cache or skip messages that were clearly safe to avoid unnecessary API usage. - **Onchain Logging + Gas Optimization** Logging every flagged message onchain had to be gas-efficient. We initially logged full messages but later optimized the contract to store just the hashes and essential metadata to reduce gas costs. - **XMTP Group Messaging Is Still Evolving** XMTP group chats are relatively new, and documentation was limited. Understanding how to reliably listen to group conversations and extract message metadata (sender, timestamp) required trial-and-error and digging through source code.