NNULLTRACE

Provider-agnostic privacy gateway for LLM traffic

Keep sensitive data out of the model context.

Add a privacy boundary between your application and any OpenAI-compatible provider. Detect and anonymize PII, protect secrets, and block prompt injection before the request leaves your boundary.

NullTrace provides technical safeguards for data minimization and security; it does not certify or guarantee your application's legal compliance.

Start free See plans
BYOKOpenAI-compatibleSelf-serve Live API status
REQUEST INSPECTORLIVE
01POST/v1/anonymize
{
"text": "Contact {{EMAIL_1}} about the invoice.",
"injection_detected": false
}
✓ PII removed before upstream delivery

SEE THE BOUNDARY

Make the privacy step visible.

These examples illustrate the transformations NullTrace is designed to perform before upstream delivery. Switching tabs is local to this page; no request is sent.

BEFORE

Contact mario.rossi@example.com about IBAN IT60X0542811101000000123456.

UPSTREAM CONTEXT

Contact {{EMAIL_1}} about IBAN {{IBAN_1}}.

Use /v1/anonymize to inspect a request before an LLM call, or /v1/chat/completions to put the boundary directly in front of an OpenAI-compatible provider.

THE CONTROL PLANE

One privacy layer, added without rewriting your stack.

01

PII anonymization

Identify emails, phone numbers, tax IDs, IBANs, cards, tokens, and secrets. Replace them with safe placeholders before inference.

02

Injection detection

Surface jailbreak and instruction override signals, with an option to block suspicious requests at the edge.

03

Provider agnostic

Keep your existing OpenAI-compatible workflow and bring your own provider credentials. NullTrace stays out of billing and model choice.

THE REQUEST FLOW

One layer between your application and your provider.

No model hosting or model marketplace: NullTrace is the middleware boundary.

01

Receive

Your application sends a request to NullTrace using your tenant API key.

02

Inspect

Supported PII, secrets, and prompt-injection signals are detected at the boundary.

03

Forward

Sanitized content is sent to your OpenAI-compatible provider with your own provider credential.

04

Return

For chat requests, NullTrace rehydrates the response before returning it to your application.

WHAT GETS PROTECTED

Concrete signals, not vague “AI safety”.

NullTrace currently ships with deterministic detection for these supported categories.

◯

Personal & financial data

Email, Italian phone numbers, Italian tax IDs, EU VAT IDs, IBANs and supported payment-card formats.

EMAILIBANIT_TAX_IDEU_VAT_ID
■

Secrets & credentials

Common high-risk credentials and tokens are replaced before they can be forwarded upstream.

JWTSTRIPEAWSPRIVATE KEY
⚠

Prompt injection

Heuristics identify instruction overrides, role manipulation, jailbreak keywords, system-prompt leakage attempts and suspicious Base64 payloads.

OVERRIDEJAILBREAKLEAKBASE64

BUILT FOR DEVELOPERS

Keep your existing provider. Change the boundary.

Start with one protected API call, then move the same layer in front of your chat traffic when you're ready.

Read API docs Get an API key
QUICK START
curl -s -X POST https://api.nulltrace.dev/v1/anonymize \
  -H "Content-Type: application/json" \
  -H "X-NullTrace-Key: <API_KEY>" \
  -d '{"text":"Contact mario.rossi@example.com","block_on_injection":false}'

Response: sanitized text + detected entities + injection signals.

SIMPLE USAGE PLANS

One privacy layer. Clear limits.

Start free and validate your integration. Paid plans are coming soon.

FREEFor testing the API

Validate your privacy boundary on a small workload before you commit.

€0forever

Includes 1,000 protected requests / month

  • PII and secret detection
  • Prompt injection signals
  • API key and usage controls
Create free account
PROFor growing products

Protect production traffic with a generous quota and the controls your team needs every day.

COMING SOON

Includes 100,000 protected requests / month

  • PII and secret detection
  • Prompt injection signals
  • API key and usage controls
Coming soon
ENTERPRISEFor high-volume teams

Protect high-volume production traffic with the complete NullTrace privacy layer and a larger monthly quota.

COMING SOON

Includes 1,000,000 protected requests / month

  • Everything in Pro
  • Higher account quota
  • Operational controls for teams
Coming soon

LIVE SERVER BENCHMARK

Measure the NullTrace engine running right now.

Run a fixed benchmark against the same core routines used by NullTrace. Results come from the current production server, so they reflect the machine and its current load.

CURRENT DEPLOYMENT

Real server-side measurements

500 iterations per routine. No network latency or upstream model time included.

Ready to benchmark this server.

—

Short payload

PII detection and anonymization on a small payload containing two sensitive entities.

—

Long payload

PII detection and anonymization across a larger payload containing multiple sensitive entities and secrets.

—

Injection scan

Prompt-injection heuristics evaluated against a representative suspicious prompt.

Results are measured inside the NullTrace release binary and can vary with server load. The benchmark does not execute arbitrary user code and does not call an upstream LLM.

LEARN & BUILD

Go deeper on the privacy boundary.

Product guides and technical resources for the problems NullTrace is built to solve.

01

PII redaction API

Detect and anonymize supported personal, financial, and secret values before an upstream LLM request.

Explore PII redaction →
02

LLM privacy gateway

Learn how the middleware boundary works while provider choice, credentials, and billing stay under your control.

Explore the gateway →
03

Prompt injection detection

See the current heuristic signals and how applications can observe or block suspicious requests.

Explore detection →
04

GDPR-oriented protection

Understand data minimization and placeholder-based transformations without implying legal certification.

Read the guide →
05

Use cases

Apply the same boundary to support tickets, AI agents, internal assistants, and document workflows.

See use cases →
06

Developer quickstart

Make one protected API call, then move the same layer in front of OpenAI-compatible chat traffic.

Open the quickstart →

Read the NullTrace technical blog →

CONTACT & SUPPORT

Need help with NullTrace?

For technical problems, use the ticketing section in your dashboard so we can keep the conversation and relevant request details together.

01

Technical support

Open your dashboard and use Support → Tickets to report bugs, API issues, account problems, billing questions, or security concerns.

Open dashboard
02

General inquiries

For general questions, feedback, or communications that do not require a support ticket, email us directly.

minerbagiorgio99@gmail.com

SECURITY & PRIVACY

Designed around a strict request boundary.

BYOK provider credentialsAPI keys stored as hashesSigned Stripe webhooksFail-closed authenticationCustomer request content not used to train modelsPublic API status checks