What is xanai?
xanai gives you access to 350+ AI models through one app and one API. You choose how you want to use them: Private runs open models inside a secure enclave, while Anonymous lets you use closed models through xanai's proxy without giving the provider your identity.
You pay for what you use with xUSD, and every answer comes with a signed receipt, batched into Merkle roots posted on BNB Chain.
Asking a question
You can ask something directly in the app or use the ask box on the homepage. Choose your model (its tier comes with it) and type your question. If you'd rather speak, tap the mic and your words will appear in the input field before you send them.
Once the model responds, you'll see a receipt below the answer with details like its hash, token count, price and the time until the next onchain root.
If you don't have enough xUSD to cover a request, you'll need to top up before sending it.
xUSD
Think of xUSD ("xanai USD") as your prepaid credit for AI.
1 xUSD equals 1 US dollar. You top up with USDT, USDC or BNB on BNB Chain. Reserves sit in an onchain vault and are public on the app's Reserves page.
You can try 3 free messages without a wallet. After that, top up in the app with any amount you need.
Each answer costs the tokens used at the model's price, plus the fee for your chosen tier. For example, 1,400 input tokens and 520 output tokens on DeepSeek V3.2 would cost around 0.000732 xUSD in this sample. The pricing page lets you check the price of any model before you ask.
xUSD is prepaid credit, not an investment. $XANAI, the planned xanai token, is not live yet and nothing in xanai is paid in it. See the token page.
Choose your tier
Private
For when you want to keep your prompts protected.
Open models like DeepSeek, Qwen, Llama, Kimi and GLM run inside a hardware enclave (TEE), a protected area of a processor. The enclave comes with an attestation you can check, and your identity is never forwarded.
The fee is 20% on top of the provider's cost.
Anonymous
For when you want to use closed models without giving the provider your identity.
Claude, GPT, Gemini, Grok and other closed models are accessed through the xanai proxy. Your identity is hidden: the provider sees your question, and only zero-data-retention endpoints are used.
Web search is available on Anonymous models only, for +0.03 xUSD per message.
The fee is 15% on top of the provider's cost.
Receipts
Every completed call gets a receipt, so you can check what happened without putting your conversation onchain.
The signed receipt records hashes for the prompt, answer, model, tier, token usage, price and timestamp. Receipts are batched into Merkle roots, which are posted on BNB Chain.
Want to check one? Head to the receipt verification page, enter the receipt hash and see its verification path and the block containing the root.
For example, a sample receipt might show 0x8f3a…c91e, DeepSeek V3.2, 1,912 tokens and a cost of 0.000732 xUSD, linked to the root 0x4c1e…9a0b. These are sample values.
The actual prompts and answers aren't stored in the onchain receipt, only their hashes.
Privacy
Your wallet address, xUSD balance and receipts are associated with your xanai account. The goal is to keep your identity separate from what you ask AI.
With Private, your prompt runs inside the enclave and your identity is never forwarded. With Anonymous, the model provider sees your question, but not your identity.
Only receipt roots are recorded onchain, without personal data or raw conversations. The site doesn't use tracking cookies, and the app uses a session cookie to keep you signed in.
Agents
Want AI inside your group chat? xanai agents can join groups on Telegram and WeChat under their own names. LINE is coming soon.
Each agent has a specific job. Market watch can share a morning brief, while Listing alerts can flag new pairs on BNB Chain, with more profiles planned.
Getting started is simple: choose an agent, connect the messenger and group, then set its daily spending limit. You can also choose the model it uses, which sets its tier.
Agents are billed per answer like any call, from your xUSD balance, within the daily spending limit you set. Each post comes with a receipt.
Agents can read and post, but they can't trade, move funds or sign transactions.
API
If you're building something of your own, xanai gives you one OpenAI- and Anthropic-compatible API for its 350+ models, at https://api.xanai.app.
Send a request to POST /v1/chat/completions, choose the model for that call (the model sets the tier), and get the response along with its receipt. The API page includes example requests, responses and routing details.
If something goes wrong, the API returns a clear error code. A 401 means there's an issue with your API key, 402 means you need more xUSD, 429 means you're sending requests too quickly, and 503 means the model isn't available.
Keep your API key on your server. Never expose it in frontend code, a public webpage or an app bundle.
Need help?
Use the Support button at the bottom of any page or in the app's sidebar to open a ticket.
If you're asking about a specific payment or request, include the receipt hash or request ID if you have it. That helps the team find the relevant call without needing your prompt or answer.
Support will never ask you to sign a transaction or send funds.
Common questions
Which tier should I use?
Choose Private when you want your prompt decrypted only inside a secure enclave. Choose Anonymous when you need a closed model and want to keep your identity separate from the provider request.
What does a receipt prove?
It lets you verify that a call was recorded with a particular model, price and timestamp. The receipt uses hashes rather than exposing your actual words.
What if the model doesn't respond?
You aren't charged if no answer comes back with a receipt.
What happens when an agent reaches its daily limit?
It stops posting until the limit resets. You can change the cap in the app whenever you need to.
Does xanai keep my prompts?
No. xanai keeps your wallet address, your balance and your receipts. A receipt holds hashes of the prompt and the answer, never the words. On Anonymous, only zero-data-retention endpoints are used. On Private, the model runs inside a TEE.
What is a TEE private LLM?
A TEE, or trusted execution environment, is a hardware enclave that keeps code and data sealed off from the machine's operator. On the Private tier, open models run inside one, with an attestation you can check, and your identity is never forwarded. The tier fee is 20%.