Skip to content
Ruvicode logo
Ruvicode
USD
Real savings vs official provider pricing

Every frontier AI model.
One API key.
Real per-request pricing.

One API key for Claude, GPT, Gemini, GLM, DeepSeek, Kimi, and more, plus image generation and editing. Pay per request, see exact costs live, and cap spending per key. Cached tokens bill at a lower rate, automatically. Top up in IDR with QRIS, or with USDC on Base.

curl https://api.ruvicode.com/v1/chat/completions \
  -H "Authorization: Bearer rvcd_..." \
  -H "Content-Type: application/json" \
  -d '{
    "model": "glm-5.2",
    "messages": [{"role": "user", "content": "Hello"}]
  }'

# Response headers include:
# X-Cost: 0.00021800
# The JSON body carries the same figure as usage.cost

AI Models

Max savings vs official

$0

Hidden Fees

700

RPM per account

One key, every model

Claude, GPT, Gemini, GLM, DeepSeek, Kimi, and more, plus image generation and editing. One key, one wallet, one bill. Switch models by changing the model id, nothing else.

Real costs, every request

Every billed response reports exactly what you paid: usage.cost in the body, streaming included, plus an X-Cost header on non-streaming calls. No credit conversion, no surprises at the end of the month.

Spend limits that hold

Set daily and monthly spend caps per key. When a limit is hit, requests stop. No runaway agent loops, no surprise bills.

Easy top-ups

Top up in IDR with QRIS from any bank or e-wallet app, on the site or in Telegram, or with USDC on Base credited automatically on-chain. Deposits from $0.01.

Works with your tools

Change one base URL and your existing OpenAI SDK, Cursor, Aider, Cline, LangChain, or agent works. No SDK swap, no code rewrite.

Rate limits that scale

Per-key rate limiting with a sliding window. Raise or lower limits per key anytime. Built for agentic workloads that burst.

Get started in 3 steps

Step 1

Create account & top up

Register with email or OAuth. Add credit with QRIS in Rupiah or USDC on Base. No commitments, no subscription.

Step 2

Get your API key

Generate a key in seconds. Set spend and rate limits per key. Your budget, your control.

Step 3

Start building

Point any OpenAI-compatible tool at our endpoint. Stream responses and see the exact cost of every request in the response.

Start building with one key.

No subscription, no lock-in. Top up what you need with QRIS or USDC and pay per request.

Get Started →

Frequently Asked Questions

How is Ruvicode cheaper than official provider pricing?

We buy inference capacity at market rates, well below what the official APIs charge, and pass most of the difference to you. The price you see on each model page is the price you pay per request. The one exception: when busy capacity forces a reroute that costs more than the listed rate, that request is billed at its real cost plus the same margin, capped at the official price. If the provider itself charged more than the official price, you pay exactly what it cost, with no margin. Every response reports its exact charge.

Do cached tokens cost less?

Yes, on most models. When a request reuses a prompt prefix that was already processed, the cached part is billed at the cached input rate shown on the model page, usually several times lower than the input rate. The gap depends on the model and moves with the market. Caching is automatic and there is nothing to enable.

Does prompt caching work with coding agents?

Yes, and agents benefit the most. Tools like OpenCode, Cline, Aider, and Cursor resend the whole conversation on every tool call, so most of each prompt is a cache hit after the first few turns. In October 2026, on the three most requested models on Ruvicode, more than 80 percent of all prompt tokens were cache reads.

How much can I save with caching?

It depends on how much of your prompt repeats and on the model's cached input rate. Repeated prefixes bill at the cached rate, unique prompts bill at the full input rate. The cost calculator lets you set the share of your prompt served from cache and shows the result for any model.

Can I see how many of my tokens were cached?

Yes. Every response's usage object reports the cache split, and your dashboard shows cached tokens per request with the percentage of the prompt they cover. Usage totals and the weekly chart include cached counts too.

Do you store my prompts or responses?

Not for chat. We do not log or store the content of your chat prompts or AI responses; only usage metadata (token counts, cost, model) is retained for billing. Image models are the one exception, by design: images you make in Studio and their prompts are kept in your private gallery so you can download them, encrypted at rest and open only to your account, then deleted automatically after 30 days (or whenever you delete them). Images from the API are kept the same way only when the request sets store to true.

Can I generate or edit images?

Yes. Image models use the same API key and wallet, through the OpenAI-compatible /v1/images/generations and /v1/images/edits endpoints, or Studio in your dashboard. They are priced per image, and you only pay for images that come back.

Do you support streaming?

Yes. Streaming works out of the box with the OpenAI-compatible API, including usage reporting in the final chunk and cached token counts. Point any OpenAI SDK at our base URL and set stream to true.

What happens if my request fails mid-stream?

You pay for what the usage object reports. A request that fails before the model produces any output costs nothing. If a stream breaks after output already reached you but before its usage arrived, you are charged an estimate for the prompt and the output delivered. If your client disconnects from a stream before any output, the prompt may still be billed as an estimate, because the provider has already processed it. When the provider later reports the real token counts, the charge is corrected to them.

Can I set spending limits per key?

Yes. Every API key can have daily and monthly spend caps. When a limit is hit, the key stops accepting requests until the next period. You can also set per-key rate limits up to 700 RPM, and every key in your account shares one account-wide rate limit, so creating more keys does not add capacity.

What payment methods do you accept?

We accept QRIS in Rupiah (IDR), paid from any bank or e-wallet app on the website or through our Telegram bot, and USDC deposits on the Base network. Top up once, pay per request. Your balance never expires.

Is it OpenAI-compatible?

Yes, for chat and images. Ruvicode serves the OpenAI Chat Completions API (/v1/chat/completions), /v1/models, and the Images API (/v1/images/generations and /v1/images/edits). Any OpenAI SDK, Cursor, Cline, OpenCode, Aider, or LangChain works by changing the base URL. The Responses API, embeddings, and audio endpoints are not available, and Anthropic-protocol clients such as Claude Code connect through 9router, a free local router that translates the protocol (the integrations page has the setup).

Do my credits expire?

No. Your wallet balance never expires. Your balance is yours to use whenever you want. No inactivity timeout, no expiry.