FOLYOO
FOLYOO GUARD

Your private way to use AI.

Let your team use ChatGPT, Claude, Gemini and more at work, without customer names, invoices or contracts ever reaching someone else's server. Folyoo Guard makes sensitive data invisible before the AI sees it.

Request accessHow it works
Works with any AIData stays in the EUGDPR & EU AI ActLive in minutes
Every field that matters for GDPR and the EU AI Act is protected and never leaves the EU.
Live detection5 sensitive fields protected
Type or paste your own text, sensitive parts are detected live.
detectedmarked by you
How it works

Mask, your AI, unmask.

Folyoo Guard is not an afterthought filter, it is a wrapper around your own AI. We replace the sensitive parts on your side, your AI works with the placeholders only, then we put the real values back locally.

01
MaskSensitive parts are swapped for realistic placeholders on your side, before anything leaves your company.
02
Your AIYour model, whichever it is (ChatGPT, Claude, Gemini or your own). It only ever sees the placeholders, with no special instructions.Never visible without Folyoo
03
UnmaskThe real values are put back locally in the answer. The model never saw them.
The problem

AI is brilliant.
Data breaches aren't.

The moment someone on your team pastes a customer name, an invoice or a contract into an AI tool, that content leaves your company and sits on someone else's servers. Under GDPR that is a real risk, and most people never even notice.

Sensitive dataNames, addresses, bank details, contracts and health data flow straight into the AI.
Foreign serversWith the big providers the data often lives outside the EU.
LiabilityIf it goes wrong, you are on the hook, not the AI provider.
The solution

We make the sensitive parts invisible.

Guard sits between your team and the AI. The delicate parts are replaced with realistic placeholders on your side. The AI only ever works with the placeholders and returns its answer. Then we put the real values back. The model never sees the real data.

01
DetectGuard finds names, companies, bank details, invoice numbers and more, in any language.
02
ReplaceEvery value becomes a realistic placeholder with a stable ID.
03
ProcessOnly the protected text goes to the AI of your choice.
04
RestoreThe real values are put back locally in the answer.
Privacy made in Germany

Built in the EU.
Compliant because nothing leaves.

Almost everyone claims to be GDPR compliant. With Folyoo Guard it is not a promise but a mechanism: sensitive content never leaves your house in the clear in the first place.

Data stays in the EUDetection can run entirely on EU servers.
GDPR compliantNo clear text, clear logs, full control.
EU AI Act readyDesigned for what is coming.
Fail-closedWhen in doubt it blocks, it never sends unprotected.
Model-agnostic

Not just ChatGPT. Any AI.

ChatGPT, Claude, Gemini, Mistral or your own model: Guard puts a protective layer in front of it. You keep the free choice, we add the privacy.

OpenAI
The GPT family, incl. the o-series.
Anthropic Claude
The Claude family for long, careful reasoning.
Google Gemini
Gemini models, strong on multimodal.
Mistral
European models, EU-hosted.
Llama & open source
Open-weight models, self-hostable.
Your own model
A private or attested (TEE) endpoint of your choice.
Pricing

Protection that scales with you.

Simple plans, final pricing on request.

Starter
Freeto try
For first tests and small teams.
1 API keyLive detectionCommunity support
Start free
Most popularTeam
On request
For teams in production.
Multiple keysEU detection endpointPriority supportAudit logs
Talk to us
Enterprise
Custom
For regulated industries.
Dedicated EU hostingYour own modelsSLA & onboardingDPA & compliance
Contact us
SECURE FROM THE GROUND UP

Fail-closed, not fail-open.

No clear text to the AIOnly placeholders leave the service. Every request is checked first.
Fail-closedIf anything stays unprotected, the request is blocked instead of sent.
EU detectionDetection can run on its own EU server, so the raw text stays in the EU.
Always the same placeholderThe same value gets the same placeholder everywhere, even across messages.
Multiple layersPatterns, AI detection and dictionaries run in parallel. Better to protect too much than too little.
Streaming & imagesLive answers and the text parts of image messages are protected too.
For your developers

Your SDK. Our domain.

No AI yet? Just use our OpenAI-compatible endpoint. Already have a setup? Keep your provider's own SDK, OpenAI, Anthropic or Google, and only point the base URL at Folyoo. The rest of your code stays the same.

OpenAI-kompatibelRecommended
import OpenAI from "openai";

new OpenAI({
  baseURL: "https://api.folyoo.com/v1",  apiKey,
});
Anthropic Claude@anthropic-ai/sdk
import Anthropic from "@anthropic-ai/sdk";

new Anthropic({
  baseURL: "https://api.folyoo.com/anthropic",  apiKey,
});
Google Gemini@google/genai
import { GoogleGenAI } from "@google/genai";

new GoogleGenAI({
  apiKey,
  httpOptions: {
    baseUrl: "https://api.folyoo.com/gemini",  },
});
FAQ

In short.

Do I have to rebuild my software?

No. If you already use an AI, you swap one address. If you have no software of your own, you get ready-made access.

Does the AI see my real data?

No, only placeholders. The real values are put back afterwards on your side, without any AI.

Does it work with my favourite model?

Yes. Guard is model-agnostic and sits in front of ChatGPT, Claude, Gemini, Mistral or your own model.

What about numbers and amounts?

Amounts stay intact by default so calculations work. Names, companies, bank details and numbers are replaced.

Your AI.
Without giving away your data.

Guard is part of the Folyoo family. Tell us which AI you use and we will set up your protected access.

Request accessWorks with any AI, live in minutes