Does the AI provider ever see my data?

No. Names, emails, and numbers are replaced with placeholders before your request leaves Andorg. The provider only ever sees masked text; the real values are restored in the answer on the way back.

What does Andorg store?

Nothing about your content. The real values, the mask mapping, and your provider API key live in RAM for one request and are then destroyed. There are no logs of your content.

How much does it cost?

Free gives unlimited chat masking at 10 requests per minute with no card. Pro is $20/mo and Max is $99/mo. You bring your own provider key and pay your provider directly — no token markup.

How hard is it to switch?

One line: point your OpenAI-compatible client's base URL at https://api.andorg.com/v1/openai and keep your own key. Works with OpenAI, Anthropic, Google, OpenRouter, and any compatible endpoint.

The AI never sees your data.
Only placeholders.

A privacy proxy between your app and your AI provider. Names, emails and numbers are masked before text leaves — and restored in the answer. Nothing is stored.

Type anything with a name, email or number — the demo runs in your browser, nothing is sent anywhere.

Built for teams that cannot leak data — fintech · healthcare · legal · agencies

    # the whole migration — bring your own key, any provider
    - client = OpenAI(api_key=KEY)
    + client = OpenAI(api_key=KEY, base_url="https://api.andorg.com/v1/openai")
  

Works with OpenAI · Anthropic · Google · OpenRouter · any OpenAI-compatible endpoint

How it works

Four steps, every request, in milliseconds.

1. Scan

Rules + a local AI model find sensitive values, on Andorg's own CPU.

hi, I'm Jane Doe
2. Mask

Each value becomes a placeholder. The mapping stays in RAM only.

hi, I'm Jane Doe hi, I'm [PERSON_1]
3. Forward

Masked text goes to your provider, on your key, at your price.

[PERSON_1] provider
4. Restore

Real values swap back into the answer. Mapping and key destroyed.

dear [PERSON_1] dear Jane Doe

Three guarantees, by architecture

Never used to train AI

Nothing is stored, so nothing can be trained on. Real values live in RAM for one request, then destroyed.

AI-breach containment

If your provider leaks, your data was never there. A breach exposes placeholders — pointers to nothing.

The model never sees you

Only masked text ever reaches the AI — on your key, at your price, from any compatible provider.

0 prompts stored 1 line of code ≤24h max data life 353 tests green

What lives where, for how long

This table is the retention policy, verbatim. Real numbers only.

Data	Where it lives	How long
Prompt text (real values)	RAM only	one request
Your provider API key	RAM only	one request
Mask ↔ value mapping	RAM only	one request
Secret-words list	encrypted at rest	until you delete
Document-scan results	encrypted at rest	≤ 24 hours
Logs of your content	nowhere	never

Pricing

Free

✓ unlimited chat masking
✓ 10 requests / min
✓ 1 API key
✓ no card required

Start free

Pro

$20/mo

✓ everything in Free
✓ 60 requests / min
✓ secret words + customization
✓ up to 10 API keys

Go Pro

for documents

Max

$99/mo

✓ everything in Pro
✓ PDF, Word, Excel, image scanning
✓ 1,000 pages / month included
✓ prepaid page packs $25

Go Max

Start free, no card. Upgrade, downgrade, or cancel anytime — you always keep the free tier.