AI Privacy Gateway

Redact sensitive data before it reaches AI models.

Zero Node detects and replaces sensitive data before your application sends it to LLMs, logs, analytics, or third-party APIs. Restore authorized values afterward without exposing the originals to the model provider.

Private local contextual models
Short-lived, single-use rehydration
Text and structured JSON
No request-content application logs

How it works

Your application keeps control of the real values.

Zero Node sits between your application and the external model. The model receives useful context, not the original sensitive data.

1

Redact

Your application sends text or JSON to Zero Node before calling an external model.

[email protected] → <EMAIL_1>
2

Call the model

The LLM works with safe, stable tokens and never needs the original customer value.

Prompt contains <EMAIL_1>
3

Rehydrate

Your authorized application restores tokens with the short-lived, single-use session.

<EMAIL_1> → authorized value

A focused privacy boundary for AI applications

Practical controls for developers who need to reduce sensitive-data exposure without breaking application workflows

🚫

No Persistent Content Store

The token↔original mapping lives in memory (or Redis) for a few minutes, never on disk, and is deleted the moment you rehydrate.

🎯

Real Detection, Not Guesswork

Email, phone, SSN, cards, IPv4/IPv6, addresses, API and URL secrets, IBAN, birth date, passport, tax, and medical IDs. Cards and IBANs receive checksum validation; contextual IDs require nearby labels.

🔁

Sanitize → Rehydrate

Redact before you call any LLM, send the tokenized text, then rehydrate the reply locally. The mapping is single-use — consumed on first rehydrate.

📈

Built to Scale Out

Sessions, rate limits, and quotas run through a shared store — point it at Redis and add replicas behind a load balancer with no code changes.

🐳

Lightweight

The API ships as a static Go binary in a minimal container, with separate private Redis and local-model services. Simple to self-host and inspect.

🔌

Simple API

Redact plain text or nested JSON, choose token, mask, remove, hash, pseudonym, or partial replacement, and rehydrate token-mode sessions once.

Try It Live

Paste synthetic text below — the application does not log it or create a reversible session. Public demo, rate-limited to prevent abuse.

No API key needed

Request

// POST /api/v1/demo/redact (no auth)
{
  "text": "Contact support at [email protected] or 555-123-4567"
}

Response

{
  "redacted": "Contact support at <EMAIL_1> or <PHONE_1>",
  "findings": [
    {"type": "EMAIL", "confidence": 0.98},
    {"type": "PHONE", "confidence": 0.85}
  ]
}

The demo uses the complete engine but creates no reversible session. Contextual PERSON, ORGANIZATION, and LOCATION detection is beta and can produce misses or false positives. Review supported formats, limits, and integration examples in the developer documentation.

Developer Quick Start

Protect a prompt with one API call.

Use the production endpoint with your beta API key. Token mode returns a short-lived session for authorized rehydration.

curl -X POST https://zero-node.com/api/v1/redact \
  -H "X-API-Key: YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "text": "Email Jane at [email protected]",
    "options": {
      "replacement_mode": "token"
    }
  }'

Simple early-access pricing

Start with one AI workflow. Beta keys are provisioned manually, and we confirm coverage and limits before activation.

Developer · Bronze

Free

  • 1,000 requests/day
  • Text and structured JSON
  • Email and phone protection
  • Public playground and email support
Try the Playground

Scale · Gold

$99/month

  • No daily key quota; abuse safeguards apply
  • Complete deterministic coverage
  • Beta person, organization, and location detection
  • Request-scoped custom patterns
  • Dedicated beta support
Request Full-Engine Trial

Plans differ by volume, support, and available coverage during beta. Review the exact detector boundaries in Detection Coverage before choosing a plan.