Your first protected AI question
Hi, I'm Kyle, the founder of FusionWall. Let's get you from a fresh installation to your first protected AI question. We will create your local account, protect the recovery code, connect the AI service you already want to use, and practice once with made-up information. Nothing important is at risk while you learn.

Create the owner account
We start with the owner account because FusionWall is private by design. This account controls the workspace on this computer. It is not a cloud profile, and your password is not sent to us. Choose a strong password. Then save the one-time recovery code somewhere separate and protected. That code is your way back in if the password is ever lost.

Protect. Review. Send.
Once you are in, the basic idea is simple. FusionWall looks for supported private details and replaces them here on your computer. You see the protected version before anything leaves. If it looks right, you approve it. When the answer returns, FusionWall restores your private values on screen and keeps a local receipt of the decision.

Connect the service you use
Now connect the AI service you plan to use. FusionWall does not include a mystery model behind the scenes. You choose the provider, the endpoint, and the exact model. Your credential is encrypted on this computer, and it never appears in the proof ledger. This connection is what brings the workspace to life while keeping the review step in front of every request.

Test the exact connection
Save the connection, then run the built-in test. We check the exact route that future questions will use. If the provider, endpoint, model, or credential changes, FusionWall asks you to test again. A passing result means the connection works. It does not mean every future AI answer will be correct, so your judgment still matters.

Ask a made-up question
For the first run, make everything up: the person, the business, and the account number. Type the question, then stop at the protected view. Read it the way you would read an email before sending. Detection is intentionally conservative, but it is not perfect. If something private is still visible, fix it. You are always the final decision-maker.

You have seen the whole boundary
That's the full first-use path. Your local account exists, the recovery code is protected, the exact AI connection has passed its test, and you have watched one safe question move through review and into a private receipt. Start small, keep reading the protected view, and never treat approval as a reflex. You are ready.