RelayVano API Relay: Claude Code Setup Without the Pairing Mistakes

The pairing mistake that breaks most Claude Code setups

Most failed Claude Code configurations are not caused by a bad key or a wrong base address. They are caused by a mismatch between the client and the model family. On RelayVano, Claude models work only in the official Claude Code and Claude Desktop clients, version 2.1.285 or newer. Pointing Claude Code at the OpenAI-compatible route instead of the Anthropic /v1/messages route will not give you the pairing you expect, and the same applies to any attempt to use a Claude model through a client that does not support it.

The same discipline applies to the other families. Codex, Cursor, Cherry Studio, Cline, OpenCode, OpenAI SDKs and cURL pair only with GPT, DeepSeek and the other non-Claude, non-Gemini models. Gemini models use the native Gemini protocol only, in Gemini CLI, and they are not OpenAI-compatible. Learning this map once removes the most common source of wasted time, because you can pick the client and the model together instead of debugging them apart.

RelayVano is an independent relay made for developers worldwide. We are not a subscription service, and we do not try to blur the boundaries between protocols. One API key reaches every model through https://api.relayvano.com, with an OpenAI-compatible /v1 route, an Anthropic /v1/messages route and a native Gemini /v1beta route. The routes are there; the pairing rules simply tell you which one your chosen client speaks.

Connecting Claude Code to RelayVano in a short setup

The setup itself is deliberately small. First, create your RelayVano account. New accounts start with $0.40 of credit, which is enough to send a test request and confirm the path works before you top up. Next, open the dashboard and copy your API key. Then configure Claude Code with the RelayVano base address, https://api.relayvano.com, and the Anthropic /v1/messages route, using a Claude model name from our catalog.

Finally, make one test request and check the usage page. You will see the request listed with its charge, so you can verify the whole chain end to end: key, route, model and cost. If you prefer to work in your own language, the whole site, dashboard and docs come in six languages: 中文, English, Español, हिन्दी, Português, Français. Prices show in yuan on the Chinese pages and in approximate dollars on the others, so a developer in São Paulo, Delhi, Madrid or Lyon can read the same numbers in a familiar format.

Two small habits make the first run smoother. Configure the Anthropic route for Claude Code rather than the OpenAI-compatible one, even though both live under the same address. And check your Claude Code version, because Claude models on RelayVano require version 2.1.285 or newer in the official Claude Code and Claude Desktop clients. Those two checks account for the majority of the friction people normally hit.

Pricing mistakes that make a cheap API look expensive

The second common mistake is comparing the wrong numbers. RelayVano charges by token, and prompt tokens, reply tokens, cache stores and cache hits are priced separately. If you ignore cache usage, your estimate drifts, because a coding session that reuses a long system prompt behaves very differently from a session that sends it fresh every time.

Compare the per-token fares directly. Claude Sonnet 5.5 is listed at $1.25 per million prompt tokens, $6.25 per million reply tokens and $0.1097 per million cache-hit tokens, against an official list price of $3 / $15. Claude Haiku 4.5 is $0.3958 prompt, $1.90 reply and $0.0319 cache hit, against an official list price of $1 / $5. Claude Fable 5.1 is $4.07 prompt, $20.33 reply and $0.3125 cache hit, against an official list price of $10 / $50. These are approximate US dollars per million tokens, with 1 USD ≈ ¥7.2, and the charge itself is made in yuan.

Here is an example, not a forecast. Say you send 200 requests a day, each with 1,000 prompt tokens and 500 reply tokens on Claude Sonnet 5.5. The prompt side is 200,000 tokens a day, which is $0.25 at the $1.25 rate. The reply side is 100,000 tokens a day, which is $0.625 at the $6.25 rate. That example totals $0.875 a day, before any separately priced cache stores or cache hits. Change the model and the arithmetic changes with it: Claude Opus 5.5 is $1.65 prompt and $7.82 reply, while MiniMax M3 is $0.1472 prompt and $0.5889 reply.

This is also where a subscription comparison needs care. RelayVano sells pay-as-you-go balance only: prepaid balance, pay per token, no monthly fee, and the balance never expires. If a vendor's subscription does not match how you actually work, a per-token relay lets you pay for the sessions you run and keep the rest.

Why every request shows its own charge

The third mistake is leaving billing in the dark. On RelayVano, every request appears on the usage page with its charge, broken down by the token categories that apply to it. You do not have to reconstruct a monthly invoice from memory, and you do not have to guess whether a long reply was billed correctly.

Interrupted streams are handled in a way that matches how real work happens. If a caller disconnects mid-stream, the request is still finished and billed on the tokens really used. A dropped connection, a closed laptop lid or a cancelled terminal command does not turn into a mystery line item. The usage page tells you what was consumed, and the prepaid balance reflects it.

Because prompt tokens, reply tokens, cache stores and cache hits are priced separately, that visibility matters more than a single headline rate. Teams that mix Claude, GPT, DeepSeek, Gemini, Qwen, Kimi, MiniMax, Hunyuan, Grok or MiMo requests can still read one usage page and see the cost of each call, in the language and currency format they chose.

Top-ups, refunds and support for a global developer base

Funding RelayVano is straightforward. Top-ups go through Stripe checkout, where cards work everywhere, plus WeChat Pay where your account supports it. Unspent top-ups can be refunded within 9 days. There is no monthly fee to cancel and no expiration clock running on your balance, so a quiet month costs nothing extra and a heavy month scales with the tokens you actually use.

RelayVano also rewards the people who bring their own teams. Referrals pay 16% of every top-up by people who sign up through your link, and 4% one level further. Earnings can be withdrawn from $5.56 or moved straight into your balance, which is useful when you want to keep working without a separate payout step.

If something does not look right, support is built into the product. Logged-in members write to us under Messages in the dashboard, so a question about a charge, a pairing issue or a model name stays attached to the account that needs help. There is no separate portal to learn and no address to hunt for.

RelayVano was built for developers worldwide, in six languages, around a simple promise: one key, transparent per-token pricing and a usage page that shows what you spent. Start with the $0.40 welcome credit, connect Claude Code through the Anthropic /v1/messages route, and send a test request. If it works as expected, top up when you are ready. Sign up for RelayVano and put your own requests on the usage page today.