Privacy

People bring this app the things they would not say out loud. That is only possible if it is careful with what it is told.

What is never stored

• The words you type. Not in logs, not in a database, not in a file.
• The words Nicodemus replies — not as text, and not as audio. A spoken answer is streamed to you and never written to disk.
• Anything that could reconstruct a conversation — no saved questions, no saved answers, no saved list of the passages a reply leaned on.

What stays only on your device

• Your conversation history, language, tradition, and access code live in your browser, never on the server. The server never receives your history.
• Your Journey progress and reflections are stored only on your device. There is no server record of what you have read.
• Story Time (a parent-opened area for young children) keeps its family profiles — a nickname and birth month — on your device only. It never uses AI to talk with children, has no links out, and nothing about a child is collected or sent anywhere.

What the server keeps — anonymous

• A random per-device id and counts (days seen, number of messages) — used for message limits and to know roughly how many people used the app.
• The shape of a conversation — for example, that a care/crisis moment occurred and its category — with no words attached. This is how we can tell whether the crisis care is working. It is anonymous, but it is not nothing, and you should know it is there.
• Token and cost totals, to know the bill.
• Audio of Scripture chapters and the fixed voice preview only — these are public and identical for everyone. Personal answers are never saved as audio.

These server-side records are aggregate, anonymous counters — usage and cost totals, and tallies of how often the care path was triggered by category — never a timestamped, per-person trail of what anyone said or did. They are kept only for the length of the pilot and cleared when its data is reset; the conversation-shape tallies are wiped on every deploy.

Which AI providers process your words

• Anthropic (the Claude model) processes the text of your questions and answers in order to generate replies.
• If you use Listen (text-to-speech) or voice input, OpenAI processes that audio — turning text into speech, or your speech into text.
• These providers process your words only to provide the feature, under their API terms. Nicodemus adds no other recipients.

Our promises

• We never sell your data. Ever.
• Every default is the private one: conversation-content logging is off, and the access gate fails closed (no code configured means no one gets in, rather than everyone).

---

This is the plain-language version. A person can, in principle, opt in to content logging only in local development — never in the live app.