A little clarity.
A lot more confidence.

Chat helps with questions, writing, ideas, and code. Build generates a single-page HTML/CSS website with a preview and source download. Images generates raster pictures from text through Hugging Face Inference Providers. Xeon AI does not currently deploy websites, provision databases, or edit an uploaded image.

A guest workspace is a quick way to explore and starts with free credits. Each new guest sign-in creates a new account, so it cannot recover an earlier guest session. Wallet sign-in uses a message signature and keeps a persistent identity. Never share a seed phrase or private key. Contract-wallet signatures are not currently supported.

Open Credits & balance and choose the free faucet. Packs contain 100,000, 250,000, or 500,000 free credits, with a maximum of 1,000,000 per account in a rolling 24-hour period. These are application credits, not blockchain tokens. The faucet is available until your account completes its first verified token top-up.

Open a completed Chat or Build result and send a follow-up. Xeon AI includes up to six recent turns from that conversation. Build revisions also include the previous source. Use New creation to start fresh. Image requests create a new image from the current prompt; previous images are not sent for editing.

The preview renders static HTML and CSS in an isolated frame. Scripts, external network resources, and navigation are blocked inside the preview. Use the mobile toggle to check a narrow layout and the Code tab to inspect the source. Downloaded code needs your own review and hosting before publication; working server logic is not included.

Choose Images, describe a scene, review the quote, and Generate. A successful request saves a real image in your account. Download keeps the original PNG, JPEG, or WebP returned by the provider. Images may not reproduce exact lettering or every detail. An unavailable provider produces an error, not a substitute demo image.

Pricing uses the task type, prompt word count, and requested feature keywords. Quick / Standard / Advanced rates are 10,000 / 20,000 / 35,000 credits across all task types. These are predictable pricing rules, not an AI assessment or measured compute cost. The quote appears before Generate and stays fixed for that request.

The quoted amount is reserved while your request is running. It is deducted once a result is successfully saved. Provider errors release the reservation without a charge. Repeating the same request after a network interruption will not create a second charge. A successful result may still need revisions, which are new priced requests.

After your first verified token deposit, your account switches to a balance denominated in the deposited token. Free credits stop being spendable and the faucet is disabled for your account only. Other accounts keep their free-credit mode. Historical credit records are preserved. The application balance shows tokens deposited into Xeon AI, not every token in your wallet.

Deposits are held in custody for your application balance. A successful AI request deducts only its quoted cost and creates one marketing-wallet allocation. Unused tokens remain available for future requests. Failed requests create no allocation. Pending settlement does not charge you again. Actual on-chain transfers remain disabled until the marketing address and custody/signing integration are configured; an allocation is not proof of a completed transfer.

That is a planned flow, but Xeon AI has not configured its token contract or a verified trading pool yet. Once supported, first obtain the correct token on the correct network, then make a separate top-up to the treasury shown on your invoice. Purchasing on a DEX alone does not credit your workspace. No purchase links are available until the contract, network, and liquidity pool are verified.

Invoices save the token address, chain, decimals, treasury, and exact amount. A worker verifies the matching ERC-20 Transfer event and finalized canonical block before crediting the deposit once. Transfers need test ETH for gas. The configured token-per-pricing-unit rate is fixed by the operator, not an exchange market price. Only standard ERC-20 tokens without transfer fees or rebasing are supported.

Cloud providers can be overloaded or exhaust their free allowance. Wait before retrying; repeated clicks do not increase the quota. Failed requests do not spend your Xeon AI balance. Adding free credits cannot increase Google or Hugging Face allowances. Free availability is not guaranteed and Xeon AI does not automatically switch to a paid provider.

Do not send the same top-up twice. If the wallet completed a transfer but the page did not save it, paste the transaction hash into the pending invoice. Verification waits for network finality, not just an initial receipt. A wrong token, amount, network, or sender may require review. Keep your transaction hash.

Chat and Build prompts and recent context go to the configured text provider; this workspace uses Google Gemini. Image prompts go to Hugging Face and its selected inference provider. Google free-tier content may be used to improve its products. Avoid confidential content. Saved prompts and results are stored by this Xeon AI instance and access is restricted to your account; they are not end-to-end encrypted.

Sessions last up to seven days and expire after 24 hours of inactivity. Disconnect revokes your current session. Switching wallet accounts signs you out so another address cannot use the existing identity. Your wallet signature signs you in; top-up transfers require a separate explicit wallet transaction.