- Do I need a credit card to start?
- No. Sign in with Google, Microsoft, or email and password, and your first 14 days run at Pro's question and connection limits (1,500 questions, 10 connections, and 8 AI schema analyses), with no payment details. After that you land on the Free plan; nothing is deleted — your connections, history, and dashboards stay, and you choose which connection stays active.
- Which currency will I be charged in?
- The one shown: USD, EUR, or GBP, at the same number in each ($18.99 = €18.99 = £18.99). The page picks a currency from your browser's timezone — the selector above the plans switches it — and checkout charges the currency you see. Once a subscription starts, its currency stays fixed.
- Can I change plans later?
- Yes. Move between Lite, Pro, and Team — and switch monthly and yearly billing — from your account settings. Teams that need invoicing can talk to sales.
- How do monthly limits work?
- Questions are metered per calendar month and reset on the 1st (UTC). Pro and Team also carry a fair-use ceiling of 500 questions per user per day. On Team, questions are pooled across seats — the whole workspace shares one monthly allowance.
- What is AI schema analysis?
- When you add a connection, Datarelix discovers what's in it — tables, columns, keys, relationships — and then the model analyses that structure so it understands your tables before you ask anything. Discovering the structure doesn't count against your allowance; only a fresh analysis does, and every plan includes a number of them per month. Resuming an analysis from a checkpoint is free. If you delete a connection, its credentials are removed immediately and the analysis is kept for 30 days — re-adding the same database restores it. What that structure changes about answer quality is covered in why semantic context matters.
- Where does my data go?
- Your database stays where it is. Queries run against it, and only the result set for the question you asked is stored, as part of the run trace. The component that runs generated code is cut off from the internet entirely, and raw query results are never passed back to the model. See the security page; the control layers behind it are documented in the read-only architecture and threat model.
- Which LLM provider do you use?
- Datarelix runs on managed frontier models — currently OpenAI, with Anthropic Claude also supported by the platform. We manage the provider and keys — there's nothing for you to configure.