The latest articles on DEV Community by T C (@trustboostai). https://dev.to/trustboostai https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F3884720%2Fb72da7c8-fefd-453a-b352-3d8acc93448d.png https://dev.to/trustboostai en T C Fri, 17 Apr 2026 15:07:56 +0000 https://dev.to/trustboostai/building-an-information-firewall-for-ai-agents-2gca https://dev.to/trustboostai/building-an-information-firewall-for-ai-agents-2gca <p>Published: true</p> <p>Description:TrustBoost is an open source middleware that redacts PII and verifies payments on Solana for autonomous AI agents.<br> tags: aiagents, opensource, privacy, solana</p> <h2> 🛡️ TrustBoost: Information Firewall for AI Agents </h2> <p>I've been building AI agents (local and cloud-based) and noticed a recurring problem: <strong>agents leak sensitive data</strong>.</p> <p>Emails, passwords, private keys, internal hostnames — all sent directly to LLMs without any control.</p> <p>So I built <strong>TrustBoost</strong>.</p> <h2> What It Does </h2> <p>TrustBoost sits between your agent and the LLM. It acts as an information firewall with three layers:</p> <div class="table-wrapper-paragraph"><table> <thead> <tr> <th>Layer</th> <th>Function</th> </tr> </thead> <tbody> <tr> <td><strong>Payment verification</strong></td> <td>Checks Solana Network transactions (149 USDC for 10,000 sanitizations) — no human approval</td> </tr> <tr> <td><strong>PII redaction</strong></td> <td>Uses GPT-4o-mini to detect and redact emails, passwords, private keys</td> </tr> <tr> <td><strong>Audit logging</strong></td> <td>Saves every operation to Google Sheets for traceability</td> </tr> </tbody> </table></div> <h2> Why Solana? </h2> <p>Agents transact with each other without humans. Solana is fast, cheap, and Helius API makes verification simple.</p> <h2> How to Test It (50 Free Requests) </h2> <div class="highlight js-code-highlight"> <pre class="highlight plaintext"><code> bash curl -X POST "https://hook.us2.make.com/h4xqu3de1qlzn9mbrf7npe8rkelpft36" \ -H "Content-Type: application/json" \ -d '{ "tx_hash": "TRIAL", "wallet_address": "your_wallet", "text": "My email is john@doe.com and my password is Secret123" }' Response: json { "status": "success", "cleaned_text": "My email is [REDACTED] and my password is [REDACTED]", "safety_score": "0.85", "risk_category": "CRITICAL" } For AI Agents (MoltBook Compatible) TrustBoost exposes a molt.json schema so agents can discover and use the service autonomously. Endpoint: https://hook.us2.make.com/h4xqu3de1qlzn9mbrf7npe8rkelpft36 Trial: "tx_hash": "TRIAL" (50 free) Paid: real Solana tx_hash (149 USDC for 10,000 sanitizations) Health check: https://raw.githubusercontent.com/teodorofodocrispin-cmyk/TrustBoost-PII-Sanitizer/main/health.json Open Source MIT licensed. Full blueprint available. 🔗 GitHub: https://bit.ly/4tOjFEj Roadmap (self-hosted version) I'm aware Make.com is not ideal for privacy-critical workloads. The next version will be a self-hostable proxy (FastAPI + Ollama) that you can run on your own infrastructure. Feedback Welcome This is a prototype. I'm building in public and learning as I go. What would make you trust this for production use? Integration with LangChain? OpenClaw? Solana payment model — yay or nay? Leave a comment or open an issue on GitHub. Thanks for reading. </code></pre> </div>