No logs
Prompts and responses pass through and are gone. We don't record IP addresses. Usage is kept only as daily totals, for billing.
Uncensored and leading open models, in your browser or plugged into your apps through the API. No email, no logs, no subscription.
Ask anything, privately… GLM-5.3 Uncensored ▾From $0.05 per 1M tokens · See pricing
Prompts and responses pass through and are gone. We don't record IP addresses. Usage is kept only as daily totals, for billing.
Your account is a random 24-character ID. No name, no email, no password to leak.
No subscription, no monthly minimum. Add credits when you need them and pay only for the tokens you use.
Uncensored
Uncensored and abliterated versions of the best open models, with the refusal training taken out. They deal with the question you actually asked, without disclaimers or moralizing. And since nothing is logged, it stays between you and the model.
Uncensored doesn't mean lawless: what you do with the answers is your responsibility.
Chat
Pick a model and start talking. Conversations are encrypted on your device before they're saved, so we can't read them.
Open chatAPI
Create a key and call any model from your code, in any language. Same balance, same prices as the chat.
POST /v1/chat Get an API keySeven currencies, the network of your choice. Monero for maximum privacy. Credits land on your balance as soon as the payment is confirmed.
Each model has a price per million tokens, for input and for output. Your balance goes down as you use it. That's the whole model.
All models and prices| Model | Input | Output | Cached |
|---|---|---|---|
| GLM-5.3 Uncensored Zhipu AI Uncensored | $0.15 | $0.24 | $0.017 |
| DeepSeek V4.1 Flash 1017 Abliterated DeepSeek Uncensored | $0.375 | $0.60 | $0.083 |
| GPT-6 Astra OpenAI | $1.25 | $2.00 | $0.083 |
| Claude Haiku 5.5 Anthropic | $0.075 | $0.12 | $0.017 |
USD per 1M tokens
No. Saved conversations are encrypted in your browser with a key derived from your account ID, which never reaches our servers. We only store ciphertext. While a model answers, your message is processed in memory, then discarded.
Mainstream models are trained to refuse whole topics. Uncensored and abliterated versions have that refusal behavior removed, so they answer directly instead of dodging. They keep the original model's abilities and are marked with a badge in the model list. You remain responsible for how you use what they write.
We can't recover it, because we never had it. Your balance and your encrypted history stay locked. Keep it in a password manager.
No request logs and no IP addresses. Billing only needs daily token totals. Our public pages load X's ad pixel to measure our ads on X; the chat, the dashboard and the Tor site don't, except a top-up's page, which tells X once it's paid. Microsoft Clarity shows us how the site is used, the chat and dashboard included but not the Tor site; your chats, account ID and keys are masked, so it never sees them. Links in our ads and posts carry a source code: we count clicks, sign-ups and payments per source, keeping nothing about you.
Yes. The API speaks both the OpenAI and the Anthropic formats, so anything that accepts a custom endpoint works: Cursor, Claude Code, Continue, Cline, Zed, Open WebUI, the OpenAI and Anthropic SDKs, your own scripts. Set our base URL, paste your API key, pick a model.
Bonus credited automatically, up to $100. No email required to sign up.