Security
SemanticGuard is built with enterprise security requirements in mind. Your data stays on your infrastructure, and every access is audited.
Authentication & Access
Multi-factor authentication (TOTP)
All users can enable 2FA via authenticator apps. Admins can require MFA for all team members.
Passkeys / WebAuthn
Phishing-resistant passwordless authentication using hardware security keys or biometrics.
Single Sign-On (SSO)
OIDC-based SSO with Google Workspace, Azure AD, Okta, and Auth0.
Role-based access control
Three roles (admin, member, viewer) with hierarchical permissions enforced on every API route.
Session management
JWT sessions with configurable expiry. Admins can revoke all sessions for any user. Role changes force re-authentication.
Data Protection
Encryption in transit
All traffic is TLS 1.2+ encrypted. HSTS enforced with preload.
Encryption at rest
Database (Neon Postgres) encrypts all data at rest. Vercel integration tokens are AES-256 encrypted before storage.
PII redaction
API keys, passwords, emails, phone numbers, and SSNs are automatically redacted from trace logs before storage. Enabled by default.
Prompt storage control
Tenant-configurable: store prompts for debugging, or disable for privacy. When disabled, only cost and cache metadata is logged.
Data retention
Configurable trace retention (1-365 days). Automatic deletion of expired logs.
Data export
Full data export via API for GDPR/CCPA compliance.
Account deletion
Complete account and data deletion via API with audit trail.
Infrastructure
Your infrastructure
SemanticGuard proxy deploys to your own Vercel account. Your databases, your cache, your data. Nothing leaves your infrastructure.
API key pass-through
Upstream API keys (OpenAI, Anthropic, Google) are passed through at request time and never stored in plaintext. Only a one-way hash is retained.
Fail-open design
If the cache layer is unavailable, requests go directly to your LLM provider. Zero downtime risk from cache failures.
IP allowlisting
Restrict API key usage to specific IP addresses or CIDR ranges.
SSRF protection
Proxy blocks requests to private IPs, metadata endpoints, and localhost.
Monitoring & Audit
Audit logging
All admin actions (key creation, settings changes, role changes, logins) are recorded with timestamps, user IDs, and IP addresses.
Request tracing
Optional per-request logging with model, cost, cache status, and latency. Configurable per tenant.
Correctness validation
Continuous automated validation of cached responses using the tenant's own AI. Failures are flagged to admins.
CSRF protection
State-changing API requests require x-requested-with header. OAuth flows use CSRF state tokens.
SOC2 Type II
In progressWe have built the technical controls SOC2 Type II requires across the Trust Service Criteria (Security, Availability, Processing Integrity, Confidentiality, Privacy). We are preparing for a formal audit; we are not yet certified.
If your procurement process requires a SOC2 report, security questionnaire, or DPA, contact our team and we will share our current control documentation and audit timeline.
Contact Security Team