Security & architecture
How NeuraVoice AI is built.
Application
NeuraVoice AI is a full-stack web application built with React and TanStack Start, served from an edge runtime. Server logic runs in server functions and server routes; the browser never holds private credentials.
Data storage
Accounts, leads, meeting requests and contact messages are stored in a managed PostgreSQL database. Row-level security is enabled on every table so signed-in customers can read and change only their own records.
Authentication
Customers sign in with email and password or Google. Private pages such as the dashboard require a valid session, and server endpoints verify the session token on every request.
AI inference workflow
When a customer requests a call brief, the browser sends the lead details to an authenticated server route. The server validates the input, calls a large language model through an AI gateway, and streams the draft back. Requests are sent with storage disabled at the model provider, and AI credentials stay on the server.
Current status
Live outbound AI calling, telephony and CRM data sync are in development. The dashboard features available today are lead management, AI call briefs and meeting requests.
Security practices
Input validation on forms and server routes, least-privilege database access, no secrets in frontend code, and search engines blocked from private pages. Security reports go to Support@neuravoiceai.com.
Compliance
We have not yet completed third-party certifications such as SOC 2 or ISO 27001. Our handling of personal data is described in the Privacy Policy and Data Processing pages.