Back to blog
Guide·May 24, 2026·5 min read

Getting started with the Claude API in Korea — switching from the OpenAI SDK, Seoul latency, and Korean tokens

A Claude API getting-started guide for Korean developers: how to switch from the OpenAI SDK with one line using base_url, where latency occurs when calling from Seoul, and Korean token usage and model selection. Model pricing and payment methods are covered separately in the Claude API pricing and payment guide.

This guide answers the questions developers in Korea most often ask about using the Claude API: pricing compared with Anthropic directly, measured latency from Seoul, whether Korean cards are accepted, and how to migrate from the OpenAI SDK with a one-line change.

First call in 5 seconds

Kunavo is 100% compatible with the OpenAI wire protocol. If you already use the OpenAI SDK, migration requires only a one-line change in base_url. Python example:

quickstart.py
from openai import OpenAI

client = OpenAI(
    api_key="sk-kn-...",
    base_url="https://api.kunavo.com/v1",  # 이 한 줄만 바꾸면 끝
)

resp = client.chat.completions.create(
    model="claude-sonnet-4-6",
    messages=[{"role": "user", "content": "한국어로 인사해주세요"}],
)
print(resp.choices[0].message.content)

It’s the same for Node.js / TypeScript:

quickstart.mjs
import OpenAI from "openai";

const client = new OpenAI({
  apiKey: process.env.KUNAVO_API_KEY!,
  baseURL: "https://api.kunavo.com/v1",
});

const resp = await client.chat.completions.create({
  model: "claude-opus-4-7",
  messages: [{ role: "user", content: "서울 오늘 날씨 알려줘" }],
});
console.log(resp.choices[0].message.content);

Sign-up is free, top up from $10, pay only for what you use (pay-as-you-go), and your balance never expires. This is enough to thoroughly test Claude Opus 4.7, Sonnet 4.6, and Haiku 4.5. The same OpenAI-compatible key works for calling not only Claude but also Gemini, GPT, and image and video models.

Price difference from Anthropic directly—Claude API costs

Claude models on Kunavo are 30% cheaper than Anthropic’s official rates. Claude Sonnet 4.6 costs $3.00 input and $15.00 output per 1M tokens directly from Anthropic, while Kunavo charges $2.10 input and $10.50 output (about 30% off). With prompt caching, cache reads cost 10% of the input rate—that is, $0.21 per 1M tokens for rereads. See Claude API pricing and payment guide and the pricing page for complete per-model prices and payment methods.

Latency from Seoul

Kunavo’s entry point is an anycast edge network, where TLS terminates at a nearby node. The gateway itself runs in a single US-East region (Ashburn, Virginia). The edge shortens the TLS handshake round trip, not the leg across the Pacific—from Seoul, round-trip time is dominated by the physical distance between Seoul and the US East Coast, plus the model’s own inference time. Calling Anthropic’s us-east-1 directly covers the same distance. Streaming is implemented as an SSE pass-through, providing a time-to-first-token on par with upstream.

Payment—domestic cards, KakaoPay?

Payments are processed by Stripe. Available methods for customers in Korea:

  • Credit / debit cards: Visa, Mastercard, American Express, JCB
  • Apple Pay, Google Pay (when configured in Chrome · Android), Link (one-click payment)
  • When checkout is displayed in Korean won: KakaoPay, Naver Pay, PAYCO, Samsung Pay, domestic cards

Prices are set in U.S. dollars ($USD), and Stripe automatically converts them to Korean won at checkout. The exchange rate includes a 2–4% conversion fee paid by the buyer; this fee does not apply when paying in dollars (Stripe Adaptive Pricing documentation, checked 2026-10-03). The minimum top-up is $10 USD (approximately 14,000 won). Because this is a prepaid wallet rather than a subscription, unused balances do not expire.

KakaoPay appears as a payment method for topping up Kunavo, while Toss is not supported. KakaoPay, Naver Pay, PAYCO, Samsung Pay, and domestic cards appear only when checkout is displayed in Korean won; domestic cards do not need to have international payments enabled on this route. When paying in dollars, you need a card that supports international payments (if international payments are blocked, enable them in your card issuer's app). These methods top up your Kunavo balance; they are not payments made directly to Anthropic.

Token costs when using Korean

Claude provides the same quality for Korean input and output as it does for English. However, Korean uses about 1.5–2 times as many tokens as English (approximately 0.5–0.7 tokens per character). Keep this in mind when calculating costs.

The most reliable way to distinguish formal and casual Korean is to specify it explicitly in the system prompt:

“You are a customer support representative in Korea. Always write every response in formal Korean (-습니다/-입니다 style), and respond to customers politely and kindly.”

Model selection guide

  • Claude Opus 4.7 — The strongest reasoning capability. Suitable for complex tasks requiring an understanding of Korean nuance, long-context summarization, and code generation.
  • Claude Sonnet 4.6 — A balance of price and capability. The model most widely adopted by Korean startups. Suitable as the main model for chatbots, content generation, and agents.
  • Claude Haiku 4.5 — Fast and inexpensive. For classification, filtering, simple summaries, and agent preprocessing. About one-tenth the cost of Sonnet.

Next steps

  1. Sign up for free—top up from $10, pay as you go, and your balance never expires
  2. Create an OpenAI-compatible API key—change only base_url to keep using your existing OpenAI SDK
  3. Make your first call with chat completions—Claude, Gemini, and GPT use the same endpoint
  4. View Claude API pricing · Payment methods explained · OpenRouter alternatives compared

Send questions to contact@kunavo.com. We’ll respond within one business day.

Related deep dives