Privacy Architecture

What Data Goes Where

Complete transparency about data flows for each AgentShield tool

💰 Token Optimizer

Marketing tool for cost analysis

📤 Data Sent to Server:

✅ Your system prompt (full text)
✅ Token count analysis
✅ Optimization suggestions

🔒 What Stays Local:

❌ Nothing - full analysis on server

Purpose:
Cost reduction analysis - NOT a security tool

🔍 Code Security Scan

Pattern-based vulnerability detection

📤 Data Sent to Server:

✅ Your code snippet (full text)
✅ Language/framework
✅ Scan type (full/quick)

🔒 What Stays Local:

❌ Nothing - analysis on server

Purpose:
Quick malware/vulnerability check - Marketing demo

🛡️ Agent Security Assessment

PRIVACY-FIRST Production Security

📤 Data Sent to Server:

✅ Agent name + platform
✅ Ed25519 public key
✅ Test scores only (passed/failed)
✅ Cryptographic signatures

🔒 What Stays Local:

✅ Your system prompts
✅ Agent source code
✅ All 77 test results (details)
✅ Vulnerability reports
✅ Private keys

Purpose:
Enterprise-grade security certification

Security Assessment Data Flow

💻

Step 1: Local Testing

77 security tests run on YOUR machine
Your prompts NEVER leave your device

🔐

Step 2: Score Submission

Only scores transmitted
"Prompt Injection: 18/18 passed"
No test payloads, no prompts

🏆

Step 3: Certificate

Ed25519 certificate issued
Valid for 90 days
Published in public registry

Why Privacy-First Architecture Matters

Traditional security audits require sending your entire system to an external service. This creates:

🔴 Data exfiltration risks (your prompts on someone else's server)
🔴 Compliance issues (GDPR, SOC2, ISO 27001)
🔴 Trust dependencies (you must trust the auditor)

AgentShield Security Assessment solves this with:

🟢 Zero-knowledge architecture (we never see your prompts)
🟢 Cryptographic proof (Ed25519 signatures)
🟢 Open-source client (audit the audit tool)

Questions About Our Privacy Architecture?

We're happy to explain our technical approach in detail.

Email Support Security Team Read Docs