3-day trial, no credit card

Stop getting rate limited.
Stop burning tokens.

Plug 191+ providers into one OpenAI-compatible endpoint. No API key yet? Use our own models with a wallet balance, or buy a ready-made plan in the Marketplace. Tokens are compressed automatically and every cost is visible in real time. Change the base URL — that is the whole integration.

191+
providers ready to use
5
API formats supported
Rp10k
per month
3 days
free trial
simulasi — perjalanan satu request lewat HematToken
AI AGENT / IDE
Claude Code · Cursor
your own app
ht-xxxxxxxx
one gateway key
HematToken
Credential vault
BYOK · AES-256
Smart routing
picks the best route
Circuit breaker
isolates failures
Budget limiter
spend limit per key
claude-sonnet-4
primary · your API key
standby
glm-4.7
backup · cheap
standby
gemini-2.5-flash
backup · OAuth
standby
→ Your AI agent or IDE sends a request — holding one gateway key and nothing else.

Your BYOK API key is safe. Stored AES-256 encrypted and decrypted only for an instant, in memory, while the request runs — never shown in plaintext, not even to our team.

Put simply: HematToken is a universal power strip with a generator built in. One socket for every device (one endpoint for every provider) — and the moment one supply cuts out (a rate limit), the backup takes over on its own. You will not notice the handover.

If any of this sounds familiar…

We built HematToken because we were tired of it ourselves. Every problem below already has an answer shipped — none of this is a roadmap promise.

429Rate limited in the middle of a session

A combo moves to another provider or another account the moment one hits its limit. You do not stop to edit a config.

Rp?The AI bill keeps growing and nobody knows why

Every request is recorded: provider, model, tokens, latency and cost. Break it down per API key and the expensive one has nowhere to hide.

🔑API keys scattered across a dozen projects

Every credential in one place, encrypted. Your application only ever holds a single gateway key.

≠Every provider wants a different request format

Send OpenAI, Anthropic, Gemini or Ollama format — we translate it into whatever the destination provider speaks.

Why it is called HematToken

Your tokens get packed first — 20–65% smaller before they reach the model

Think of tokens as airline baggage: providers charge you by the kilo. Token Saver is the person who packs it — every request goes through four compression stages before it flies. Same contents, 20–65% less of them.

RTK

Drop the duplicates — repeated tool output, diffs and logs are trimmed.

Headroom

Fold what folds — JSON and tables are compacted, losslessly.

Caveman

Tighten the prose — verbose sentences get dense, meaning intact.

Ponytail

Tie off the tail — long conversation history is summarised.

Anything that cannot be reconstructed exactly is left alone. The model's answer is the same; only the token count moved, and all of it is recorded in your usage dashboard.

simulasi — satu request lewat pipeline Token Saver
Ukuran konteks request10.000 tokens
1
RTK
Drop the duplicates — repeated tool output, diffs and logs are trimmed.
2
Headroom
Fold what folds — JSON and tables are compacted, losslessly.
3
Caveman
Tighten the prose — verbose sentences get dense, meaning intact.
4
Ponytail
Tie off the tail — long conversation history is summarised.
On its way to the model…
Typical saving20–65%tokens per session

Three steps, done

Nothing to install, no server to run. It is all already up in the cloud.

1

Sign up & connect a provider

Paste your provider API key, or sign in over OAuth where the provider supports it. Every credential is stored encrypted.

2

Create a gateway key

One key for your application. Restrict it to certain providers or models, and give it a spend limit.

3

Change the base URL

Point your application at the HematToken endpoint. The request format stays exactly what you use today.

Two lines change. That is the migration.
from openai import OpenAI

client = OpenAI(
    base_url="https://api.hemattoken.id/v1",
    api_key="ht-xxxxxxxxxxxx",
)

client.chat.completions.create(
    model="anthropic/claude-sonnet-4",
    messages=[{"role": "user", "content": "Halo!"}],
)

What you get

All of it works today. Nothing here is "coming soon".

One endpoint, many providers

Point your application's base URL at HematToken and pick a model with "provider/model". 166 API-key providers, 16 over OAuth, and 94 of them have a free tier.

Combos & auto-fallback

Build a list of backup models with a fallback, round-robin or fusion strategy. When one provider goes down or hits a limit, the request continues to the next.

Token Saver

Lossless context compression: tool output, diffs and logs are tidied before they reach the model. Anything that cannot be reconstructed exactly is left untouched.

Real-time usage & cost

Requests, tokens, latency and cost — per provider, per model and per API key. Including a ranking of free models by how much you actually use them.

Quota Tracker

Watch what is left of your Claude, Codex and other subscriptions on one screen, with the time each one resets.

Guardrails

Detect PII and prompt injection before a request leaves. Redact it, block it, or just record that it happened.

Spend limit per key

Put a cost and token ceiling on each gateway key. Safe to hand to a teammate or drop into CI.

Ready-made plans in the Marketplace

No provider account yet? Buy a plan from a partner we have vetted — once it is paid it appears in Providers as an ordinary connection, quota and validity included.

More than chat

Embeddings, web search, page fetch, text-to-image, text-to-speech and transcription — all through the same gateway.

191+ provider, tiga cara nyambung

The ones people reach for most are below — the rest are waiting in the dashboard.

OAUTH16 providers— sign in directly, no API key to copy and paste
Logo Claude CodeClaude CodeLogo OpenAI CodexOpenAI CodexLogo GitHub CopilotGitHub CopilotLogo Cursor IDECursor IDELogo WindsurfWindsurfLogo Kilo CodeKilo CodeLogo ClineClineLogo KimiKimiLogo xAI GrokxAI GrokLogo ZedZed+6 lainnya
FREE TIER94 of 191— start without paying a provider anything
Logo NVIDIA NIMNVIDIA NIMLogo OpenRouterOpenRouterLogo GeminiGeminiLogo CloudflareCloudflareLogo Ollama CloudOllama CloudLogo Vertex AIVertex AILogo Kiro AIKiro AILogo OpenCode FreeOpenCode Free+86 lainnya
API KEY166 providers— paste your own key and go
Logo OpenAIOpenAILogo AnthropicAnthropicLogo DeepSeekDeepSeekLogo GroqGroqLogo MistralMistralLogo Together AITogether AILogo CerebrasCerebrasLogo Fireworks AIFireworks AILogo Azure OpenAIAzure OpenAILogo Amazon BedrockAmazon BedrockLogo HuggingFaceHuggingFaceLogo PerplexityPerplexity+154 lainnya

Provider not listed? Add any custom endpoint that speaks OpenAI or Anthropic.

Works with the tools you already use

No special SDK. Point the base URL at us, use your gateway key, done.

Logo Claude Code
Claude Code
Anthropic CLI
Logo OpenAI Codex
OpenAI Codex
Codex CLI
Logo Cursor
Cursor
AI code editor
Logo Cline
Cline
coding agent for VS Code
Logo GitHub Copilot
GitHub Copilot
AI pair programmer
Logo Kilo Code
Kilo Code
coding agent
Logo Roo Code
Roo Code
coding agent
Logo Continue
Continue
AI assistant in your IDE
Logo Windsurf
Windsurf
AI code editor
Logo Zed
Zed
very fast editor
Logo Trae
Trae
AI IDE
Logo Factory Droid
Factory Droid
autonomous agent

…and endpoints compatible with the SDK you already have:

OpenAI SDK & Chat Completions
/v1/chat/completions
Claude Code & Anthropic SDK
/v1/messages
Codex & OpenAI Agents SDK
/v1/responses
Google GenAI SDK
/v1beta/models/…
Ollama-compatible tools
/api/chat
No API key? No problem

Use HematToken's own models — top up, pay per token

Besides bring-your-own-key, we host a selection of models you can use with no API key and no account to connect. Top up your wallet and pay only for the tokens you use — each model is served by several providers at once and requests are routed between them automatically.

Top up the wallet

Add balance in Rupiah through local payment methods. No extra subscription — the balance is drawn down as you use it.

Use it straight away, no key

Hosted models appear in your model list like any BYOK model — usable in combos, compressed by Token Saver, recorded in usage.

Routed automatically

One model can be served by several providers behind the scenes. One slow or down? The request moves by itself — nothing for you to think about.

Models available
16 models · price per 1M tokens
DeepSeek V4.1 Flash
1 provider · 1M konteks
$0.18 / $0.72
DeepSeek V4 Pro
1 provider · 1M konteks
$0.792 / $2.38
MiniMax M2.7
1 provider · 197K konteks
$0.36 / $1.44
MiniMax M3
1 provider · 524K konteks
$0.72 / $2.88
GLM 5.1
1 provider · 203K konteks
$1.2 / $3.84
GLM 5.2
1 provider · 1M konteks
$1.68 / $5.28
Kimi K2.6
1 provider · 262K konteks
$1.14 / $4.8
Kimi K2.7 Code
1 provider · 262K konteks
$1.14 / $4.8
Kimi K3
1 provider · 1M konteks
$3.6 / $18
GLM 5.3
1 provider · 1M konteks
$1.44 / $4.53
GLM 5 Turbo
1 provider · 200K konteks
$1.24 / $4.12
GLM 4.7
1 provider · 200K konteks
$0.618 / $2.27
GLM 4.6
1 provider · 200K konteks
$0.618 / $2.27
GLM 4.5
1 provider · 128K konteks
$0.618 / $2.27
GLM 5.3 Flash
2 provider · 1M konteks
$0.18 / $0.6
Qwen3.5
1 provider · 262K konteks
$0.72 / $4.32
The list and prices above come from the same catalogue as the Models menu. Balance empty? Requests to hosted models are simply declined — there is no quiet bill.
Top up & try itSee subscription pricing
Or buy a ready-made plan

Partner plans that arrive as a connection

As well as a per-token wallet, you can buy a quota plan from a partner we have vetted. Pay once, see the quota up front, and the moment it settles the plan shows up in Providers as an ordinary connection — it joins combos, fallback, Token Saver and usage like any other.

  • Paid once, not a subscription — nothing renews on its own.
  • Quota, validity and rate limits are visible before you pay.
  • We handle the key; you never sign up with the partner.
Available in the Marketplace now
9 plans · one-off price
★ Pilihan
Zrouter
DeepSeek V4 Flash · 346M
Rp 24.500Rp 34.000-27%
Quota
346M tokens
Validity
30 days
Reset
one-shot
RPM
not stated
TPM
not stated
Concurrent
not stated
deepseek-v4-flash
★ Pilihan
BandelBanget
BandelBanget 150M
Rp 31.000Rp 45.000-31%
Quota
150M tokens
Validity
28 days
Reset
one-shot
RPM
not stated
TPM
not stated
Concurrent
not stated
deepseek-v4.1-flash(2.56x)deepseek-v4.1-mod(3.2x)deepseek-v4-flash(2x)deepseek-v4-flash-0731(2x)deepseek-v4-flash-vision-exp(2.5x)deepseek-v4-mod(3x)deepseek-v4-pro(1.15x)deepseek-v4-pro-0813(1.8x)glm-5.1glm-5.2(1.25x)glm-5.2-mod(3x)glm-5.3(2x)glm-5.3-flash(2x)glm-5.3-flashx(2.5x)glm-5.3-mod(3.2x)hy3kimi-k2.6kimi-k2.7-codekimi-k2.7-code-highspeed(1.5x)kimi-k3(2x)kimi-k3-mod(3.5x)mimo-v2.5-pro(1.4x)minimax-m3(1.4x)
★ Pilihan
SmartsProxy
deepseek-v4.1-flash 100M (Vision)
Rp 46.000Rp 72.500-36%
Quota
100M tokens
Validity
28 days
Reset
one-shot
RPM
not stated
TPM
not stated
Concurrent
not stated
deepseek-v4.1-flash
★ Pilihan
AIHack
Starter
Rp 47.000Rp 50.000-6%
Quota
200M tokens
Validity
10 days
Reset
one-shot
RPM
not stated
TPM
not stated
Concurrent
not stated
claude-opus-4.5claude-sonnet-4.6
★ Pilihan
SnifoxAI
Quantum B
Rp 80.000Rp 90.000-11%
Quota
490M tokens
Validity
1 day
Reset
one-shot
RPM
30 req/min
TPM
not stated
Concurrent
not stated
claude-opus-4.7(1.55x)claude-opus-4.8(1.55x)claude-opus-5(1.8x)claude-sonnet-4.6(1.5x)claude-sonnet-5(1.55x)gpt-5.5(1.45x)gpt-5.6-sol(1.7x)gpt-5.6-terra(1.45x)gpt-6-luna(1.03x)
SmartsProxy
Token MOCIN 1M [TESTER]
Rp 1.900Rp 2.500-24%
Quota
1M tokens
Validity
1 day
Reset
one-shot
RPM
not stated
TPM
not stated
Concurrent
not stated
kimi-k3deepseek-v4-flash-0731deepseek-v4-pro-0813minimax-m3kimi-k2.7-code-highspeedglm-5.1glm-5.2kimi-k2.7-codemimo-v2.5-progemini-3.7-flashdeepseek-v4-flash-vision-expdeepseek-v4-flashdeepseek-v4-prodeepseek-v4-modhy3glm-5.3-flashkimi-k3-modglm-5.3-moddeepseek-v4.1-flashdeepseek-v4.1-modglm-5.3hy4qwen3.8-2.4t-a95bqwen3.8-27bqwen3.8-flashqwen3.8-maxqwen3.8-max-0902qwen3.8-livetranslate-flash-realtimeqwen3.8-omni-flash
BandelBanget
BandelBanget 10M
Rp 3.000
Quota
10M tokens
Validity
28 days
Reset
one-shot
RPM
not stated
TPM
not stated
Concurrent
not stated
deepseek-v4.1-flash(2.56x)deepseek-v4.1-mod(3.2x)deepseek-v4-flash(2x)deepseek-v4-flash-0731(2x)deepseek-v4-flash-vision-exp(2.5x)deepseek-v4-mod(3x)deepseek-v4-pro(1.15x)deepseek-v4-pro-0813(1.8x)glm-5.1glm-5.2(1.25x)glm-5.2-mod(3x)glm-5.3(2x)glm-5.3-flash(2x)glm-5.3-flashx(2.5x)glm-5.3-mod(3.2x)hy3kimi-k2.6kimi-k2.7-codekimi-k2.7-code-highspeed(1.5x)kimi-k3(2x)kimi-k3-mod(3.5x)mimo-v2.5-pro(1.4x)minimax-m3(1.4x)
BandelBanget
BandelBanget 25M
Rp 6.000Rp 7.500-20%
Quota
25M tokens
Validity
28 days
Reset
one-shot
RPM
not stated
TPM
not stated
Concurrent
not stated
deepseek-v4.1-flash(2.56x)deepseek-v4.1-mod(3.2x)deepseek-v4-flash(2x)deepseek-v4-flash-0731(2x)deepseek-v4-flash-vision-exp(2.5x)deepseek-v4-mod(3x)deepseek-v4-pro(1.15x)deepseek-v4-pro-0813(1.8x)glm-5.1glm-5.2(1.25x)glm-5.2-mod(3x)glm-5.3(2x)glm-5.3-flash(2x)glm-5.3-flashx(2.5x)glm-5.3-mod(3.2x)hy3kimi-k2.6kimi-k2.7-codekimi-k2.7-code-highspeed(1.5x)kimi-k3(2x)kimi-k3-mod(3.5x)mimo-v2.5-pro(1.4x)minimax-m3(1.4x)
AIHack
Daily
Rp 7.400Rp 10.000-26%
Quota
20M tokens
Validity
1 day
Reset
one-shot
RPM
not stated
TPM
not stated
Concurrent
not stated
claude-opus-4.5claude-sonnet-4.6
Prices above include the admin fee and come straight from the Marketplace catalogue. Stock and the plan list change — only what can be bought right now is shown.
See every plan

Pricing that makes sense

The subscription unlocks every feature. How you pay for model calls is your choice: bring your own provider API key, top up a balance for the models we host, or buy a ready-made plan from a Marketplace partner.

Start here
TRIAL
Free/3 days

Everything unlocked — no credit card, no feature gates. It starts the moment your email is verified.

Try free for 3 days
MONTHLY
Rp10.000/month
Rp20.000save 50%
Termasuk

Bundled with BandelBanget 10M Plan

  • + BandelBanget 10M from BandelBangetRp2.000 Gratis10M tokens · valid for 28 days from activationBandelBanget
Start the 3-day trial
QUARTERLY
Rp30.000/3 months
Rp60.000save 50%
Termasuk

Bundled with BandelBanget 25M Plan

  • + BandelBanget 25M from BandelBangetRp5.000 Gratis25M tokens · valid for 28 days from activationBandelBanget
Start the 3-day trial
6-MONTH
Rp60.000/6 months
Rp120.000save 50%
Termasuk

Bundled with BandelBanget 50M Plan

  • + BandelBanget 50M from BandelBangetRp10.000 Gratis50M tokens · valid for 28 days from activationBandelBanget
Start the 3-day trial
Best value
YEARLY
Rp110.000/year
Rp240.000save 54%

Ten months' price for twelve — two months cheaper than paying monthly.

Termasuk

Bundled with Quantum C Plan and BandelBanget 25M Plan

  • + Quantum C from SnifoxAIRp30.000 Gratis21M tokens · valid for 3 days from activation · resets daily · 30 req/minSnifoxAI
  • + BandelBanget 25M from BandelBangetRp5.000 Gratis25M tokens · valid for 28 days from activationBandelBanget
Start the 3-day trial
Every subscription gets exactly the same features. Only the billing period differs.
Access to 191+ providers
Unlimited combos & auto-fallback
Token Saver & compression
Usage, cost and quota tracking
Guardrails & spend limits
Embeddings, search, image, TTS, STT
Webhooks
Pay-as-you-go balance for our models
Marketplace partner plans
Invite a friend, earn credit

Every friend who subscribes with your code earns you Rp5,000 of credit toward your own renewal. There is no cap on how many.

Your friend saves too: 20% off their first subscription, on the monthly plan or the yearly one.

Get started

Pay by scanning a QRIS code

no credit card · no manual transfer

Subscriptions, Marketplace plans and wallet top-ups — all of it is paid through QRIS. Scan with any bank or e-wallet app that supports it, confirm, done. Thanks to Bank Indonesia's QRIS Cross-Border scheme, if you are in a partner country you can pay with your own country's payment app.

Countries already connected to QRIS

  • Indonesiahome
  • Malaysia
  • Singapore
  • Thailand
  • Japan
  • South Korea
  • China
  • India & Saudi Arabia coming

The list follows Bank Indonesia's QRIS Cross-Border expansion and can grow.

What's new

Feature releases, price changes, and what we are working on. Anything you have not opened is marked — the mark lives in this browser, so there is nothing to sign in to.

13 Sept 2026

Pentingnya Guardrails dan Mematuhi ToS

Kami Tidak Akan Berkompromi!!

1 Sept 2026

Response Injection

Kami selalu menjaga dan menghindarkan user terkait response injection

31 Aug 2026

HematToken X BandelBanget

kolaborasi mitra provider baru

30 Aug 2026

Privasi dan Kebijakan

kami tidak menyimpan prompt dan balasan dari AI kamu

28 Aug 2026

Soft Launching

Soft Launching dan promo menarik

Questions people actually ask

Not necessarily. You can bring your own key (paste an API key or sign in with a provider OAuth account — we handle routing, fallback and the savings), or use the models HematToken hosts: top up the wallet and pay per token used, with no key at all. Both work side by side.

Ready to start?

Three days free, no credit card. Your first provider can be connected in minutes.

Sign up freeAlready have an account? Sign in
Support