The trust layer for AI applications

Risk, compliance, and abuse protection for production AI — before it becomes a liability.

Get started free View pricing

Every modern stack has layers for critical concerns. AI trust is missing.

Stripe → payments · Sentry → errors · PostHog → analytics · SignalVault → AI trust

The problem

AI moved to production. Trust didn't.

Logging is an afterthought

Teams ship AI features fast. Audit logs come later—if at all.

Silent data leaks

PII and secrets slip through prompts. You won't know until it's too late.

No audit trail

When things break, you have no record of what was sent to the AI.

Compliance after the fact

Security teams ask for logs after an incident. You have nothing to show.

How it works

Drop-in protection that works with your existing AI stack

Architecture

Your App → SignalVault → LLM Provider
1

Route requests

AI requests flow through SignalVault's proxy

2

Log securely

Every prompt and response is encrypted and stored

3

Detect violations

Rules catch PII, secrets, and policy breaches in real-time

4

Get alerts

Instant notifications and audit-ready exports

Built for production AI

Solve real problems teams face when shipping AI to customers

Stop data leaks

Automatically detect and block PII, API keys, and secrets before they reach AI providers.

Prove compliance

Immutable audit logs ready for SOC2, GDPR, and security reviews. Export anytime.

Control costs

Track token usage, set budgets per user, and catch abuse before your bill explodes.

Battle-tested in production at RepurposeAPI

Built for teams shipping AI features to real customers

Early access — we'll help you integrate

Frequently asked questions

What is SignalVault?
SignalVault is an infrastructure layer that sits between your application and AI providers like OpenAI. It logs every AI interaction with an encrypted audit trail, enforces guardrail rules (PII detection, secret scanning, token limits), and provides compliance exports for SOC2 and GDPR.
How does SignalVault detect PII in prompts?
SignalVault uses pattern-based detection (regex) to identify emails, phone numbers, SSNs, credit card numbers, and API keys in prompts before they reach the AI provider. You can configure rules to block, warn, or redact when PII or secrets are detected. When redact is configured, matched content is replaced with [REDACTED:pii] or [REDACTED:secret] before storage — the original text never reaches the database.
Does SignalVault slow down AI requests?
SignalVault adds minimal latency. Rule evaluation happens in-process before the request is forwarded. For teams that need zero added latency, Mirror Mode logs requests asynchronously without sitting in the request path.
What AI providers does SignalVault work with?
SignalVault works with any AI provider that uses the OpenAI-compatible API format, including OpenAI, Anthropic (via compatible wrappers), Azure OpenAI, and self-hosted models. Our SDKs wrap the OpenAI client library.
Is my data encrypted?
Yes. All prompts and responses are encrypted at rest using AES-256-GCM via Cloak. API keys are hashed with HMAC-SHA256 before storage. All connections use TLS. Raw keys and unencrypted payloads are never persisted.
How does pricing work?
SignalVault uses per-app pricing. Each app you monitor has its own subscription: Starter at $49/month (30-day retention, production only) or Growth at $149/month (90-day retention, custom rules, exports, alerts). Enterprise plans are available for teams with custom requirements.