Claude Code on Relay365: pay-as-you-go relay service

The first key, bought by tokens instead of seats

The Relay365 duty desk opens every shift the same way: one person, one account, one key, and a side project that needs a model now and then. There is no seat to buy and no monthly fee. Relay365 is an independent third-party relay run like a round-the-clock duty desk, so budgeting a weekend build means budgeting tokens instead of counting colleagues who might log in.

Signing up puts $0.25 of signup credit on the wallet. From there the desk hands over a key you can tag and bind to a lane, plus quick-wiring snippets for the API base. Because the wallet is prepaid, the first request is also the first decision: which lane, and roughly how many tokens the project will send each day?

That is the shift log this article follows. It starts with the first key and ends with the first bill, and along the way it shows how a small project can keep every lane, every token type and every charge on one table instead of spreading them across separate dashboards.

Reading the shift log of tokens

On Relay365, the four token types are priced separately: input, output, cache write and cache read. That separation is what makes a token budget legible. You can see which side of the request actually costs, and you can adjust the project accordingly rather than guessing at a monthly subscription.

For example, if you send 1,000,000 input tokens and 100,000 output tokens a day to Claude Fable 5, the input side is $3.79 and the output side is $1.894, so about $5.68 a day. The same shape of work on Qwen3.7 Plus costs far less in the example above: 1,000,000 input tokens plus 500,000 output tokens comes to $0.15 plus $0.2243, or about $0.37 a day. These are examples, not quotes, but the arithmetic comes straight from the Relay365 rate table.

Cache traffic gets its own line. Claude Fable 5 cache read is $0.2917 per 1 million tokens, so 5,000,000 cache read tokens in a day would add about $1.46, while a cache write on the same lane is $4.66 per 1 million tokens. Four separate prices mean four separate levers, and the desk can tell you which one you pulled.

Every request can be traced in the usage log, which shows the latest 200 requests with the time, model, input, output, cache recall, charge and result of each call. One honest detail: the log does not show which key made the call, so if you run several keys, keep your own naming tidy and let Relay365 handle the token accounting.

Picking a lane that matches the client

Client rules are firm here, and the desk would rather set expectations early. Claude Fable 5 works only from the official Claude Code and Claude Desktop clients, with Claude Code 2.1.285 or newer. In Cursor, Cherry Studio, Cline, OpenCode, an SDK or cURL, use the GPT, DeepSeek, Kimi, GLM, Qwen, Grok, MiniMax or Hunyuan lanes instead.

Gemini takes its own route as well. Gemini 2.5 Flash runs on its native protocol only, through Gemini CLI: single replies use :generateContent, streams use :streamGenerateContent?alt=sse, and requests cannot go through /v1/chat/completions. For example, 1,000,000 input tokens plus 200,000 output tokens on Gemini 2.5 Flash would be $0.2389 plus $0.382, about $0.62, with cache write at $0.2389 and cache read at $0.0236 per 1 million tokens.

Where the rate table quotes an official list price, the desk can show the gap. GPT-6 Astra output is $3.72 per 1 million tokens on Relay365, about 7% of the official list price of $50, and its cache read is $0.075 against an official $1. For Claude Fable 5, the output rate of $18.94 is about 38% of the official $50, and cache read of $0.2917 sits against an official $1. These comparisons cover only the families where a list price is quoted.

When a duty pass covers the night shift

Sometimes a project is not a trickle but a sprint, and that is where Relay365 duty passes fit. A pass comes with a dedicated key that never touches the wallet, and it covers Claude or GPT lanes. Other models always draw on the wallet.

The Claude pass for 14 days costs $50.97, draws up to $390 of quota at the duty-pass rate x2.5, and allows up to $210 of quota per 7 days. In list-price terms that is at most about $156 of usage, and less on models whose pass rate sits above list. The GPT pass for 30 days costs $81.53, draws up to $2350 of quota at the duty-pass rate x2.5, and allows up to $700 of quota per 7 days, which is at most about $940 of usage at list prices.

A call already let through on a pass is drawn in full from the pass even if it runs past the quota or the 7-day ceiling, and the overrun counts against the earliest-expiring card. When a pass is nearly used up, each running call holds $3 x the draw rate, and an extra parallel call gets HTTP 429 until the running ones finish. A side project can therefore plan a sprint without quietly draining the wallet it uses for everyday work.

The first bill and what follows it

Here is how the first bill arrives. The wallet is prepaid, there is no monthly fee, and every call is settled on the token counts the upstream reports. A call that fails upstream with a 502 or a 503 is not billed. A disconnected stream is billed for what was actually used, which is exactly the kind of line the desk wants a small project to be able to read and check.

The ledger in the dashboard keeps wallet, bounty and pass accounts side by side, and in-site messages reach the desk once you log in. Six interface languages are available. There are also guardrails: HTTP 429 appears when a single key sends more than 120 calls in 60 seconds, or when an account keeps more than 8 requests in flight.

Top-ups are refundable within 14 days: unused balance comes back minus what was consumed and the payment channel's fees, while bounty and promotional credit are not refundable. The referral bounty pays 18% at level 1 and 6% at level 2 of what referred members pay, and it can be moved into the wallet or withdrawn from $11.11.

So the shift closes the way it opened. One key, one prepaid wallet, one table showing every lane and every token type, and a usage log that lets you trace each request from the first call to the first bill. When the next side project is ready, create a Relay365 account and let the duty desk take the first request.