DEV Community

TrustBoost-PII-Sanitizer
TrustBoost-PII-Sanitizer

Posted on • Edited on

Building an Information Firewall for AI Agents

Published: true
Description: TrustBoost is an open source middleware that redacts PII and verifies payments on Solana for autonomous AI agents.
tags: aiagents, opensource, privacy, solana.

⚠️ UPDATE May 20, 2026 — TrustBoost v2.6.0 is live.

New in v2.6.0:
→ Proof of Sanitization on Solana — every paid sanitization
is anchored on-chain via Helius. Verifiable by anyone:
GET api.trustboost.dev/verify/{anchor_tx}
→ 8 languages verified in production: EN, ES-LATAM,
PT-BR, DE, JA, FR, IT, KO
→ Live demo (no registration):
huggingface.co/spaces/TrustBoost/pii-sanitizer
→ Launching on Product Hunt today

⚠️ UPDATE May 18, 2026 — TrustBoost v2.5.0 is live.

New in v2.5.0:
→ Context-Aware Sanitization — field context in /sanitize
accepts: legal/financial/medical/code/general
→ Privacy Budget per Agent — operators configure daily
limits once, agents operate autonomously within them
→ TrustBoost Score — GET /score/{wallet} for M2M trust
verification: TRUSTED/VERIFIED/ACTIVE/NEW tiers
→ MCP Server: api.trustboost.dev/mcp
Claude Code · Cursor · Windsurf · Glama compatible
→ GitHub Action: github.com/marketplace/actions/trustboost-pii-scanner
→ 8 languages: EN, ES-LATAM, PT-BR, DE, JA, FR, IT, KO
400+ installs on ClawHub. All examples below remain valid.

⚠️ UPDATE April 27, 2026 — TrustBoost v2.0 is live.
Migrated from Make.com to FastAPI + Supabase + Render (AWS).
New endpoint: https://api.trustboost.dev/sanitize
Open source server: github.com/teodorofodocrispin-cmyk/trustboost-api
All examples below have been updated to reflect v2.0.

🛡️ TrustBoost: Information Firewall for AI Agents

I've been building AI agents (local and cloud-based) and noticed a recurring problem: agents leak sensitive data.
Emails, passwords, private keys, internal hostnames — all sent directly to LLMs without any control.
So I built TrustBoost.

What It Does

TrustBoost sits between your agent and the LLM. It acts as an information firewall with three layers:

Layer | Function
Payment verification | Checks Solana Network transactions (149 USDC for 10,000 sanitizations) — no human approval
PII redaction | Uses GPT-4o-mini to detect and redact emails, passwords, private keys, national IDs, crypto wallets
Audit logging | Saves every operation to Supabase PostgreSQL for traceability

Why Solana?

Agents transact with each other without humans. Solana is fast, cheap, and Helius API makes verification simple. And now — every paid sanitization is anchored on-chain. Proof that cannot be falsified.

How to Test It (50 Free Requests Per Wallet)

curl -X POST "https://api.trustboost.dev/sanitize" \
-H "Content-Type: application/json" \
-d '{
"tx_hash": "TRIAL",
"wallet_address": "your_wallet",
"text": "My email is john@doe.com and my AWS key is AKIAIOSFODNN7EXAMPLE"
}'

Response:

{
"status": "success",
"data": {
"sanitized_content": "My email is [REDACTED] and my AWS key is [REDACTED]",
"safety_score": 0.85,
"risk_category": "CRITICAL",
"context_applied": "general",
"usage_metrics": {
"quota_remaining": 49,
"quota_limit": 50
}
}
}

Proof of Sanitization on Solana (NEW in v2.6.0)

Every paid sanitization returns an immutable on-chain proof:

{
"proof_of_sanitization": {
"solana_tx": "abc123...",
"verify_url": "https://solscan.io/tx/abc123...",
"description": "Immutable proof anchored on Solana mainnet"
}
}

Verify any sanitization independently:
GET https://api.trustboost.dev/verify/{anchor_tx}

Context-Aware Sanitization (NEW in v2.5.0)

curl -X POST "https://api.trustboost.dev/sanitize" \
-H "Content-Type: application/json" \
-d '{
"tx_hash": "TRIAL",
"wallet_address": "your_wallet",
"text": "RFC: LOPJ850101ABC, monto: $50,000",
"context": "financial"
}'

Multilingual Support — 8 Languages

🇺🇸 English — SSN, API keys, credit cards, passwords
🇲🇽🇨🇴 Spanish LATAM — RFC, CUIT, DNI, CURP, Cédula
🇧🇷🇵🇹 Portuguese — CPF, CNPJ, RG, NIF
🇩🇪 German — Personalausweis, Steuernummer, IBAN DE
🇯🇵 Japanese — マイナンバー, 運転免許証, 住所
🇫🇷🇧🇪🇨🇦 French — NIR (Sécurité Sociale), SIRET, Carte Vitale, IBAN FR
🇮🇹 Italian — Codice Fiscale, Partita IVA, Tessera Sanitaria
🇰🇷 Korean — 주민등록번호 (RRN), 사업자등록번호, 여권번호

For AI Agents (MCP + MoltBook Compatible)

MCP Server: https://api.trustboost.dev/mcp
Endpoint: https://api.trustboost.dev/sanitize
Health: https://api.trustboost.dev/health
Verify: https://api.trustboost.dev/verify/{anchor_tx}
Trial: tx_hash=TRIAL — 50 free requests per wallet_address
Paid: 149 USDC → 10,000 sanitizations + on-chain proof

v2.6.0 Infrastructure — Open Source

Component | Technology
API Framework | FastAPI (Python)
Database | Supabase PostgreSQL
Hosting | Render (AWS)
Payment Oracle | Helius on Solana
Privacy Engine | OpenAI GPT-4o-mini
MCP Server | JSON-RPC 2.0
Proof Layer | Solana via Helius Memo transactions

Full server code auditable at:
github.com/teodorofodocrispin-cmyk/trustboost-api

Important Limitations

Not suitable for HIPAA zero-transmission environments
Payments are irreversible — test with TRIAL before paying
TRIAL is trust-based — paid mode uses on-chain verification
Prototype stage — not yet SOC2 certified

Open Source

MIT licensed. Full source code available.
🔗 GitHub: github.com/teodorofodocrispin-cmyk/TrustBoost-PII-Sanitizer
🔗 Server: github.com/teodorofodocrispin-cmyk/trustboost-api
🔗 Live Demo: huggingface.co/spaces/TrustBoost/pii-sanitizer
🔗 ClawHub: clawhub.ai/teodorofodocrispin-cmyk/trustboost-pii-sanitizer
🔗 Glama: glama.ai/mcp/servers/teodorofodocrispin-cmyk/trustboost-api
🔗 GitHub Action: github.com/marketplace/actions/trustboost-pii-scanner
🔗 Product Hunt: producthunt.com/products/trustboost-pii-sanitizer

Feedback Welcome

EU AI Act enforcement is August 2, 2026 — 74 days away.
What would make you trust TrustBoost for production compliance?
Leave a comment or open an issue on GitHub.
Thanks for reading.

Top comments (0)