None of the "private" or "anonymous" platforms that let you access LLMs were really that private or anon. Nor did they have that many features beyond a basic chat. So, I built it with Nymbot. The service is incredibly private, can be completely anonymous, and has robust features matching or exceeding other offerings. Just a shortlist: - No account email, phone number or password: A Nostr key made on your device is your identity, your history and your balance. Or, you can use your existing Nostr identity. - End-to-end encrypted: every question and answer travels as a gift-wrapped Nostr event, readable only by you and Nymbot. - Quantum-resistant: ML-KEM is combined with the classical key exchange, so traffic captured today can't be read later. - Anonymous mode: move a chat onto a throwaway key and spend credits moved across as blind vouchers, so a reply can't be traced back to you. - Paid in Bitcoin: no subscription and no card. You buy credits over Lightning and pay per reply: base 10 sats for a standard reply, 100 for a frontier model. - Free daily replies: a free allowance of standard replies each day, before you buy anything. Actually free, no data-mining. - Every frontier model: Claude, GPT, Gemini, Grok, Kimi, Qwen and MiniMax. You can pin one per chat, or let Nymbot pick a model for each message. - Compare models: ask two models the same question side by side. - Works in your repo: connect GitHub, GitLab or Gitea. Nymbot reads and searches your code and, with writes turned on, commits, branches and opens pull requests, asking you first each time. - Run code: Python and JavaScript run free on your own device, and optional paid server runs handle anything bigger. This works in shared chats too. - Artifacts and live preview: HTML and other code in a reply can be previewed in a locked-down frame. - Image and video: more than twenty generator mdoels. You can also edit a photo you attach, or ask what's in a picture. - Web search and deep research: web access adds context for answers and for sourced research with a Pro model. - Documents: attach large PDFs and Nymbot searches them, citing the pages it used. - Teams: split a big task across 2 to 4 parallel workers. - Connectors: plug in MCP tool servers per chat. - Memory: standing facts about you, carried between chats, that you can read and delete. - Personas, bots and workspaces: custom instructions, bots you write yourself, and ready-made chat setups. - Scheduled prompts: have Nymbot ask something on a schedule and notify you. - Voice: dictation, and replies read aloud in the reply's language. - Encrypted share links: share a read-only chat whose key stays in the link, with an optional post to Nostr. - Many conversations: unlimited chats with folders, tags, pins, archive, search, forks and export. - Sync everywhere: encrypted settings and chats follow your key across devices. Your identity and balance are shared with Nymchat. - Key backup: back up your key with a passkey, or with a PIN through Apple or Google. - Gifts and transfers: send credits as a gift link, or move your balance to another key. - Every platform: a web app you can install, plus Android and iOS apps. https://nymbot.ai

Replies (3)

.'s avatar
. 3 weeks ago
I am trying to understand what you are saying. As I understand user prompts are still readable by the upstream AI if not in an attested TEE.
↑