VELTANDocs

Point your tools
at one endpoint.

Explore the VELTAN buyer flow in the compute console. The SDK example below describes the API preview.

import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://api.veltan.xyz/v1",
  apiKey: process.env.VELTAN_API_KEY,
});

const response = await client.chat.completions.create({
  model: "anthropic/claude-sonnet",
  messages: [{ role: "user", content: "Review this code." }],
});

// Interface example. The API backend is not live.

Start with compute

1. Choose a model from the example catalog.
2. Open the console and review a compute allocation.
3. Create a demo key and copy it while it is shown once.
4. Run a simulated request and inspect usage and settlement.

API STATUS / PREVIEW. https://api.veltan.xyz/v1 is the configured preview endpoint. Hosted authentication and model providers are not connected.

Authentication.

Create a named demo key in the console. The complete secret is shown once and can be copied before you close the dialog.

1. Open the demo console and enter a key name.
2. Select CREATE DEMO KEY and copy the displayed demo secret.
3. Close the dialog. Only the masked prefix and key metadata remain.
4. Use REVOKE to change its local status to revoked.

API preview request header

Authorization: Bearer $VELTAN_API_KEY

Demo key metadata is stored in this browser. Hosted authentication, secure key storage, scopes and rate limits are not connected.

Exact model routing.

One interface can support multiple model identifiers. The model you request is the model the future gateway will route, without silent substitution.

The current catalog has 7 stable demo entries. Search by model or provider, filter by category and select TRY MODEL to preselect that model in the console.

Identifiers

anthropic/claude-opus
anthropic/claude-sonnet
openai/gpt
google/gemini
xai/grok
deepseek/chat
moonshot/kimi

Reference and VELTAN input/output prices are illustrative USD rates per one million tokens. They are not verified provider prices or a live discount commitment.

Chat completions.

The VELTAN API preview follows the familiar chat-completions request structure. Use an exact model identifier and an array of messages.

Request fields

import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://api.veltan.xyz/v1",
  apiKey: process.env.VELTAN_API_KEY,
});

const response = await client.chat.completions.create({
  model: "anthropic/claude-sonnet",
  messages: [{ role: "user", content: "Review this code." }],
});

// Interface example. The API backend is not live.

In the demo console, a request produces a deterministic sample output. The selected token counts are used to quote and meter local usage. No request is sent to an AI provider.

Fund a cap.
Meter usage.

A compute allocation defines a spend cap. Every simulated request consumes part of the available balance. Unused value stays available as compute balance.

Allocation quote

The shared getComputeQuote(faceValue) service calculates the example funding amount before an allocation is reviewed.

faceValue = desired compute value
savings = faceValue × example discount rate
fundingRequired = faceValue − savings

Request metering

The inference quote combines input and output token counts with the selected model's example rates. Requests exceeding the available balance are rejected locally. Funding simulations increase the demo cap and available balance and create an allocation ledger entry.

Unused value remains compute balance. The demo does not send a USDC refund to a wallet.

Meter every
request.

The console tracks total requests, generated tokens, compute used and compute available. Each simulated request appends a local usage record and updates the current balance.

Account summary

Model breakdown

Claude, GPT, Gemini, Grok and DeepSeek totals are deterministic example data. Selected-model simulations update the matching model breakdown. The request table is a sample of account activity, not an exhaustive list of seeded requests.

Account state and allocation records persist in this browser. They do not represent provider billing or onchain activity.

Settle what
you use.

A settlement record connects a compute window, its cap, metered usage and the unused value. Inspect these fields in the console's settlement view.

Verifiable billing

The intended product uses auditable settlement records. Real chain verification and explorer links require a confirmed transaction from the future Solana adapter.

Demo transactions are not onchain. No wallet signature, transfer or network verification occurs in this phase.

USDC on Solana.

The funding asset is USDC and the target settlement network is Solana. Current allocations are browser-local simulations.

1. Select ADD COMPUTE in the console.
2. Choose a preset or enter a custom compute value.
3. Inspect face value, savings and the example funding requirement.
4. Select REVIEW ALLOCATION.
5. Select SIMULATE FUNDING to update the local compute balance.

Reserved adapter interfaces

connectWallet()
getUSDCBalance()
getTokenBalance()
createFundingIntent()
verifyFundingTransaction()
getSettlementTransaction()

Future supported wallets are Phantom, Solflare, Backpack. Wallet connections, mainnet transfers, transaction signing and real USDC payments are not active.

The app never asks for a seed phrase or private key.