AI API for Claude Code, OpenCode & agents
No 5-hour limit.One key, many models.
Stop waiting for quota resets in the middle of your work. Quota is counted in tokens, not hours, and you can switch models any time without a new account. Pay with QRIS or USDT, no credit card needed.
OpenAI & Anthropic compatible · base_urlhttps://api.tokenomixai.com/v1
- Claude Code · OpenCode · Hermes · 9router
- TOKENOMIXAIauth · limit · route · meter
- Upstream modelsSSE ↩
- 01tools
Keep the tools you already use
Claude Code (terminal and VS Code), OpenCode, 9router, Hermes or the OpenAI SDK. Just change the base URL and key.
- 02multi-model
Switch models, not accounts
GLM, DeepSeek and Kimi have stable stock, GPT and Claude are available when in stock. Switching models is a one-line change.
- 03qris · usdt
No credit card needed
Pay with QRIS from an Indonesian bank app or e-wallet, or with USDT. No subscription, no automatic billing.
- 04rollover
Leftover tokens never expire
Top up any time with your account hashtag, even with the smallest package. Leftover tokens are merged and the validity is renewed.
01AI Token Pricelist
Ready-to-use AI tokens, no subscription.
02Integration
Change base_url. Done.
OPENAI_BASE_URL=https://api.tokenomixai.com/v1
ANTHROPIC_BASE_URL=https://api.tokenomixai.com
Quickstart: install the tool + set your key, 1 command
Detailsbash <(curl -fsSL https://tokenomixai.com/install/opencode.sh)- 1. Open Terminal, paste the command above, then press Enter.
- 2. When asked, paste your API key (sk-tkx-...). The key is not shown while you type, that is normal. Do not share it with anyone.
- 3. Open a new terminal, then run
opencode.
Your old config (if any) is backed up automatically before it is replaced.
Want to go back to the original config?bash <(curl -fsSL https://tokenomixai.com/install/opencode.sh) --restore
New model not showing in /models yet?bash <(curl -fsSL https://tokenomixai.com/install/opencode.sh) --update-models
curl https://api.tokenomixai.com/v1/chat/completions \
-H "Authorization: Bearer $OPENAI_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "MODEL_ID",
"stream": true,
"messages": [{"role": "user", "content": "Hello!"}]
}'03Features
Infrastructure you can inspect.
- 01/v1/messages
OpenAI and Anthropic formats
One base URL for the OpenAI SDK, Claude Code and agents. Streaming and tool calling work in both formats.
- 02macOS · Linux · Windows
One-command install
Installs Claude Code or OpenCode and fills in the base URL and key. Your old config is backed up and can be restored.
- 03GET /v1/quota
Remaining quota where you work
Remaining quota shows under the Claude Code input box, and OpenCode warns you at 10% left. Your own scripts can call /v1/quota.
- 04up to 5 keys
A separate key per tool
Create up to 5 extra keys for your laptop, server or bot. They share one quota and one rate limit. If one leaks, revoke just that one.
- 05404 model_not_found
Typos do not cost quota
A model name that is not in the catalog or a malformed parameter is rejected at the gateway, before it can use any quota.
- 06600 req/min
Generous rate limit
600 requests per minute per account, shared by all your keys, with no concurrency cap at the gateway. Over the limit you get a 429 with Retry-After.
- 07metadata only
Prompts are not stored
The gateway only logs time, model, status, latency and token count. Prompt and response content is not stored.
- 08SHA-256
Keys stored as a hash
Once you pick up your key, it is only stored as a SHA-256 hash. Rotating it in the Portal disables the old key right away.
- 09#TKX-…
Top up without touching config
Top up on the website or the Telegram bot with your account hashtag. The key and config in your tools stay the same.
04Models
Live catalog, open multipliers.
tokens_used = (prompt + completion) × multiplier
Price /1M tokens = normal price per token × multiplier, before promos. The low end is for the largest package, the high end for the smallest: the bigger the package, the cheaper each token. One command in a coding tool can become dozens of API calls, and each call resends the whole conversation context.
Availability: GLM, DeepSeek and Kimi have stable stock. GPT and Claude are available when in stock, and their status shows live in this table.
Having trouble? Contact the admin via WhatsApp or Telegram and describe the issue. We will help until it is solved.
05Calculator
Estimate first, then buy.
tokens/day = requests × tokens × multiplier
= 100 × 4,000 × 1×
Usage estimate
400Ktokens / day
A rough estimate based on your input. Actual usage depends on prompt and response length.
06FAQ
Frequently asked questions.
How do I pay?
Pick a package, enter your email (new customer) or account hashtag (top-up), then pay with QRIS from any Indonesian bank app or e-wallet (the amount is already filled in), or with USDT via Binance UID, Bybit UID, BSC, TRC20 or Solana. Once the payment comes through, your API key or top-up appears automatically on the order page. Prices are in Indonesian Rupiah (IDR); for USDT the checkout converts them at the rate shown there. You can also order through our Telegram bot @TokenomixAIBot with /paket.
Can I use Claude Code without a Claude Pro subscription?
Yes. Point Claude Code at your TokenomixAI API key, no Claude Pro account and no credit card needed. You pick the model from the catalog, such as GLM, DeepSeek and Kimi, plus Claude when in stock. The key is yours alone, not an account shared with other people, and quota is counted in tokens, not a 5-hour window.
Is TokenomixAI legit and safe?
You can check what we claim before you pay: the API status and model catalog on this page are live from our server, the installer scripts can be read before you run them, and the Telegram group is open for questions without buying. You pay with QRIS or USDT for an amount shown before you pay, the key arrives once the payment lands, and refund rules are written in the Refund Policy (in Indonesian). If you are unsure, start with the smallest package.
How long is a package valid, and do leftover tokens expire?
Every package is valid for 28 days. Leftover tokens do not expire: when you top up with the same account hashtag, the remaining tokens are merged with the new ones and the validity follows the latest top-up.
What is a multiplier?
Every model has a token multiplier. A model with a 2× multiplier deducts 2,000 tokens for a 1,000-token task. Each model's multiplier is shown in the model catalog.
Are my prompts stored?
The TokenomixAI gateway only logs request metadata: model, status, latency and token count. Prompt and response content is not stored by the gateway. Requests are still processed by the upstream model provider to produce the answer.
Does it work with Claude Code, OpenCode, Hermes or 9router?
Yes. Claude Code and OpenCode have a one-command install that also fills in the base URL and key. Claude Code uses the Anthropic format with the base URL https://api.tokenomixai.com (without /v1). Other tools with an OpenAI-compatible mode (custom provider) only need the TokenomixAI base URL and API key. Per-tool guides are on the Documentation page.
How do I check my remaining quota?
Open the Portal and sign in with your API key. If you used the installer, remaining quota also shows under the Claude Code input box, and OpenCode notifies you at 10% left. For your own scripts, call GET /v1/quota with the same key. If you bought through the Telegram bot, you can also use /cek.
My key leaked or I lost it. What now?
If your main key leaked, sign in to the Portal and rotate it. The old key stops working immediately. If an extra key leaked, revoke just that key in the Portal. If you lost your key, contact us with your order code and we will help.
Contact
Need help?
Ask directly.
Orders, payments, high-volume needs or tool setup. We reply on WhatsApp and Telegram.
An open group for setup questions, workflow sharing and updates. No purchase needed. Mostly in Indonesian.