Show HN: Veila – AI chat through privacy proxy with end-to-end encrypted history

2 points by fjulian 3 months ago · 0 comments · 2 min read

Hi HN — I built Veila, a privacy‑first AI chat service. I wanted something that prevents model providers from profiling users and linking information from chats to their identity.

I'm a robotics engineer by training, this is my first public launch of a web app.

Try it: https://app.veila.ai (free tier, no email required)

  - What it is:
    - Anonymous AI chat via a privacy proxy (provider sees our server, not your IP or account info)
    - End‑to‑end encrypted history, keys derived from password and never leave your device
    - Pay‑as‑you‑go; switch models mid‑chat (OpenAI now; Claude, Gemini and others planned)
    - Practical UX: sort chats into folders, Markdown, copyable code blocks, mobile‑friendly

  - Notes/limits:
    - Not self‑hosted: prompts go to third‑party APIs
    - If you include identifying info, upstream sees it
    - Prompts take a bit long sometimes, because reasoning is set to "medium" for now. Plan to make this adjustable in the future.

  - Looking for feedback:
    - What do you need to trust this? Open source? Independent audit?
    - Gaps in the threat model I'm missing
    - Which UI features and AI models you'd want next
    - Any UX rough edges (esp. mobile)

Compare Veila to ChatGPT, Claude, Gemini, etc. (best viewed on desktop): https://veila.ai/docs/compare.html

Homepage: https://veila.ai

Discord: https://discord.gg/RcrbZ25ytb

More background: https://veila.ai/about.html

Happy to answer your questions here.

No comments yet.