Use the API, pay with x402, or run Kronos yourself

Three ways to get Kronos forecasts into your agent. Use our API, install the free skill, or run the model yourself. Pick the one that fits.

Live today: BTC-USD and ETH-USD, 1h candles. Coming soon: stocks, ETFs, forex. Payments: live in USDC on Base and Solana mainnet. Live payment activity: /stats.

1. Use the API

No signup and no API key. Start free, then pay per call when you need the latest forecast.

Symbols: BTC-USD, ETH-USD. Timeframe: 1h. Machine-readable docs: /openapi.json · /llms.txt · /llms-full.txt · /.well-known/x402

Try it now:

curl "https://quantx402.app/api/v1/sample?symbol=BTC-USD&tf=1h"

# The paid endpoint answers 402 + a PAYMENT-REQUIRED header (base64 JSON) until you pay:
curl -i "https://quantx402.app/api/v1/forecast?symbol=BTC-USD&tf=1h"

You're never charged for errors. Unsupported symbols, missing forecasts, and forecasts that are 2+ candles old return 4xx/5xx and the payment isn't settled.

2. Pay per call with x402

When you call /api/v1/forecast without paying, it answers HTTP 402 with a PAYMENT-REQUIRED header: the price (20000 USDC base units = $0.02), the accepted networks, and the pay-to address. Your client signs a USDC payment, retries with a PAYMENT-SIGNATURE header, and gets 200 with the JSON plus a PAYMENT-RESPONSE receipt.

// npm i @x402/fetch @x402/core @x402/evm viem
// Pays real USDC on Base mainnet: use a dedicated agent wallet with a small balance
import { wrapFetchWithPayment, x402Client } from "@x402/fetch";
import { ExactEvmScheme } from "@x402/evm/exact/client";
import { privateKeyToAccount } from "viem/accounts";

const client = new x402Client();
client.register("eip155:8453", new ExactEvmScheme(privateKeyToAccount(process.env.EVM_PRIVATE_KEY)));
const payFetch = wrapFetchWithPayment(fetch, client);

const res = await payFetch("https://quantx402.app/api/v1/forecast?symbol=BTC-USD&tf=1h");
const facts = await res.json();
console.log(facts.direction, facts.prob_up, facts.band_80_pct, facts.invalidation_level);

Payments run on Base mainnet (eip155:8453) and Solana mainnet (solana:5eykt4UsFv8P8NJdTREpY1vzqKqZKvdp). To pay from Solana, also register ExactSvmScheme (@x402/svm/exact/client) with a Solana signer for that network. The step-by-step post walks through every request and header.

Good habits: give the agent wallet a small balance, check the decoded price before paying, and keep keys in environment variables, never in code or prompts.

3. Install the free skill

FDWA Kronos agent skill (MIT). It gives an agent the instructions and a script to run Kronos locally on any OHLCV data and return the same fact-list format the API uses. You need no account, and the guide covers the payment side.

Get the free skill →

4. Run Kronos yourself

Kronos is open source (MIT) by its original authors: paper, code, weights on Hugging Face. QuantX402 isn't affiliated with them. It runs on a CPU, and a GPU makes it faster.

git clone https://github.com/shiyu-coder/Kronos && cd Kronos
python -m venv .venv && source .venv/bin/activate
pip install torch pandas einops huggingface_hub safetensors

from model import Kronos, KronosTokenizer, KronosPredictor
tok = KronosTokenizer.from_pretrained("NeoQuasar/Kronos-Tokenizer-base")
mdl = Kronos.from_pretrained("NeoQuasar/Kronos-small")
p = KronosPredictor(mdl, tok, device="cpu", max_context=512)
pred = p.predict(df, x_timestamp, y_timestamp, pred_len=24, T=1.0, top_p=0.9, sample_count=1)

Pitfalls we hit in testing:

• sample_count averages the paths and returns one line, so you lose the spread. To keep all paths, repeat the series N times in predict_batch with sample_count=1.
• Use closed candles only, and pass no NaNs. You have to build the future timestamps yourself. For stocks, that means using a market calendar so weekends and overnight hours are skipped.
• Settings change answers. Model size, lookback and temperature can flip the direction. Compare two setups and report it when they disagree.

Running it yourself is free. The API exists for when you'd rather not maintain PyTorch, weights, data feeds, a scheduler and the interpretation layer.

5. Go deeper: Kronos for AI Agents

This guide is free. For the full walkthrough, the PDF Kronos for AI Agents — Teach Your Trading Bot to Read Candles (19 pages, $47 on Gumroad) covers, per its product page: setup in 5 minutes, multi-path forecasts, Alpaca paper trading, agent fact lists, data storage, a sentiment add-on, and Kalshi & Polymarket mapping.

Building a wider AI agent workforce? FDWA's other playbook, AI Agents Money Lab (21 pages, $47), reviews 8 agent systems and includes setups with code and 5 money-making automations.

Educational material, not financial advice. Neither guide promises returns.

QuantX402 forecasts are probabilistic model output for research and education. Not financial advice. Past results do not guarantee future results. Payments are live in USDC on Base and Solana mainnet.