Classmethod Malaysia · Managed LLM gateway

One API key for the best AI models, billed in your own currency.

Kakehashi puts the leading AI models behind a single OpenAI-compatible endpoint. Top up with a payment method you already use, in the currency you already hold, get an invoice with your local tax on it, and spend that credit across every model. No cloud contract, no US-dollar card, no monthly minimum.

One endpoint
every model in the catalogue
Your currency
local payment methods, tax on the invoice
Prepaid
spend only what you have topped up

Who it is for

Built for the teams the big providers make wait

Getting frontier models in Southeast Asia usually means a US entity, a US card and a contract nobody will sign for a pilot. Kakehashi exists to remove all three.

  • Teams across Southeast Asia

    You want the same models a San Francisco startup uses, but you are paying from a local bank account and your finance team needs an invoice in your own currency, with your own tax on it.

  • Developers who already use the OpenAI SDK

    You do not want a second client library or a per-vendor contract each time a better model ships. One base URL, one key, the model chosen per request.

  • Anyone who cannot sign an open-ended bill

    Agencies, student teams, side projects and pilots: you need a ceiling you set yourself, not a card quietly absorbing whatever a runaway loop costs.

Why Kakehashi

Frontier models, on terms a team here can actually sign

The models are the easy part. What usually stops a team is the billing, the paperwork and the fear of an open-ended invoice — so that is what Kakehashi fixes.

  • One endpoint, every model

    Point the OpenAI SDK at one base URL and choose the model per request. Swapping models is a string change, not a second integration — and the Claude Code CLI takes the same endpoint.

  • Pay in your own currency

    Top up with a card or one of the local payment methods your country already uses, charged in the currency you hold. Local tax is shown before you pay and itemised on the receipt.

  • Prepaid, with a hard ceiling

    Credit is bought up front and shared across your keys, and each key can carry its own budget cap. Nothing can quietly run up a bill you never agreed to.

  • See the spend as it happens

    Per-key usage, daily spend and the full top-up history live in the portal — with email or Slack alerts at the thresholds you choose.

  • Your prompts are not kept

    Request and response bodies are never persisted after the response is returned to you. What is recorded is the usage a bill is made of: tokens, cost, model, timestamp.

  • Built for a team, not just a laptop

    Invite colleagues into one organization, split them into groups, and let the whole team draw on a single pool of credit with per-group keys and budgets.

Models

Pick the model per request, not per contract

Pass the name in the request body and the gateway routes it. This is the live catalogue; prices and anything added since are on the portal's model page once you are signed in.

Anthropic

  • claude-haiku-4-5200K context
  • claude-sonnet-4-61M context
  • claude-opus-4-6-v11M context

DeepSeek

  • deepseek-v3.2164K context
  • deepseek-r1128K context

Moonshot AI

  • kimi-k31M context

OpenAI

  • gpt-oss-120b128K context

Qwen

  • qwen3-next-80b256K context
  • qwen3-vl-235b256K context
  • qwen3-coder-next256K context

Quick start

Your first request is three lines of change

Kakehashi speaks the OpenAI API. If your code already calls an OpenAI-compatible endpoint, the integration is a base URL and a key.

  1. Sign in

    Your invite arrives by email. Sign in with the email link or with Google — there is no password to invent.

  2. Issue a key

    Create a key in the portal and give it a budget. Revoke or re-issue it any time; the credit stays with your account, not the key.

  3. Change the base URL

    Keep the OpenAI SDK you already use. Swap in this base URL and your key, and the model name decides the rest.

curl https://kakehashi-litellm-proxy-goe6aori3q-as.a.run.app/v1/chat/completions \
  -H "Authorization: Bearer $KAKEHASHI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "claude-sonnet-4-6", "messages": [{"role": "user", "content": "Hello"}]}'

Replace $KAKEHASHI_API_KEY with the key the portal issues you.

Credits

Top up what you need. Spend it on any model.

There is no subscription and no monthly minimum. You buy credit up front, it is charged in your own currency at the day's rate, and usage is drawn from it until it runs out.

  • List price plus a small usage fee

    Each request costs the model's own per-token rate plus a usage fee. The portal quotes every model fee-inclusive, so the number you plan against is the number you pay.

  • Tax on the top-up, not the request

    Your local tax is charged once, when you buy credit, and it is itemised before you confirm the payment — nothing is added per call afterwards.

  • Credit is yours until you spend it

    Balances are not reset at the end of the month and the pool is shared across your keys, so a quiet month simply carries forward.

Start with a top-up, not a procurement cycle.

Kakehashi is invite-only while we are in early access. Tell us what you are building and we will open an account — usually the same working day.