Relay365 API relay: Claude Code key checks for a new account

Your first Relay365 hour: sign up and read the key table

Welcome to the Relay365 duty desk. Relay365 is an independent third-party relay, built to be run like a round-the-clock duty desk, and the first hour after you sign up should be about orientation rather than guesswork. Create your account, and a new account starts with $0.25 in signup credit. From there, the dashboard is the place to settle three questions quickly: which keys exist, which lane each key is bound to, and how a single request is billed.

The dashboard gives you API keys with tags, a key bound to a lane, quick-wiring snippets, a usage log of the latest 200 requests, a ledger of wallet, bounty and pass accounts, in-site messages to the desk (log in first), and six interface languages. That is a compact control surface, but it is enough to answer the questions that matter early on, and every lane sits on one table so you can compare the model families you are considering without jumping between pages.

Billing is straightforward to explain. Relay365 uses a prepaid wallet with no monthly fee, and every call is settled on the token counts the upstream reports. A call that fails upstream with a 502 or 503 is not billed. A disconnected stream is billed for what was actually used. Keep those two rules in mind while you run your first test requests, because they define what a small error costs.

Tag the key, bind the lane, and check the client rules

Tags are the simplest habit to start with. When you create a key, attach a tag that tells you what the key is for in your own bookkeeping, such as the project or environment it serves. A key is also bound to a lane, and a lane is a model family. Binding a key to a lane keeps your intent explicit: the key is not a blank pass, it points at one family of models.

One honest caveat belongs in this first hour. The usage log shows the time, model, input, output, cache recall, charge and result of each of the latest 200 requests, but it does not show which key made the call. So tags are for your own record keeping and hygiene, while the log is where you trace traffic and charges across the service. That combination is useful: you know what each key was meant for, and you can still follow every request that went through Relay365.

Next, confirm the client rules before you wire anything up, because they are firm. Claude Fable 5 works only from the official Claude Code and Claude Desktop clients, and Claude Code 2.1.285 or newer is required. Cursor, Cherry Studio, Cline, OpenCode, any SDK or cURL pair only with GPT, DeepSeek, Kimi, GLM, Qwen, Grok, MiniMax and Hunyuan lanes.

Gemini 2.5 Flash follows its own native protocol only, through Gemini CLI. It uses :generateContent for single replies and :streamGenerateContent?alt=sse for streams. It is not OpenAI-compatible and cannot go through /v1/chat/completions, so plan the endpoint before you plan the snippet.

Finally, note the request limits, because they shape how you test. HTTP 429 appears when there are more than 120 calls in 60 seconds on one key, or more than 8 requests in flight for one account. If you see a 429, you are pushing one key or one account too hard, and the fix is pacing rather than retrying in a tight loop.

Trace the first request in the usage log

Send a small test request, then open the usage log and read the row. You should see the time, the model, input, output, cache recall, the charge and the result. That single row is the fastest way to confirm that the lane, the client and the billing all agree with each other.

Relay365 prices four token types separately: input, output, cache write and cache read. The rates are per 1 million tokens in approximate USD, while the account itself is kept in CNY. For Claude Fable 5, input is $3.79, output $18.94, cache write $4.66 and cache read $0.2917. Its output rate is about 38% of official list price, where official output is $50. For example, if you send 1,000,000 input tokens and 1,000,000 output tokens of Claude Fable 5 a day, that is $3.79 plus $18.94, or $22.73 a day, before you add any cache write or cache read traffic.

GPT-6 Astra is priced differently, and the pricing page shows the full table. Its input rate is $0.7444, output $3.72 and cache read $0.075, and its output rate is about 7% of official list price, where official output is $50. DeepSeek V4 Flash lists input $0.2889, output $1.15 and cache read $0.006944, with no cache write rate shown. No official price is quoted for that family, so we make no comparison there.

Gemini 2.5 Flash lists input $0.2389, output $1.91, cache write $0.2389 and cache read $0.0236. Again, no official price is quoted for the family, so the honest move is to read the numbers as they are and pick the lane that fits your workload.

If a key leaks, retire it and rotate

A leaked key is a desk emergency, and the first-hour checklist should already tell you where to act. Retire the exposed key in the dashboard, create a replacement, bind the new key to the same lane, tag it with the same label so your own records stay clean, and update the quick-wiring snippets or environment variables that referenced the old key. Then send one small test request and confirm it appears in the usage log.

While you rotate, keep the billing rules close. A call that fails upstream with a 502 or 503 is not billed, while a disconnected stream is billed for what was actually used. If you were throttled into a 429, that is a pacing signal, not a billing event to argue about.

If something looks off, the dashboard includes in-site messages to the desk, with a sign-in required. Send the request time, the model, the lane and the result you saw, and the duty desk can pick it up from there.

Confirm the wallet, the duty pass and the refund window

Open the ledger and confirm your wallet, bounty and pass accounts separately. The referral bounty pays 18% at level 1 and 6% at level 2 of what referred members pay, and it can be withdrawn from $11.11 or moved into the wallet. Refunds are available within 14 days of a top-up, with unused balance refunded minus what was consumed and the payment channel's fees. Bounty and promotional credit are not refundable, so keep that distinction clear when you top up.

Two duty passes are worth a look in the first hour, because each comes with a dedicated key that never touches the wallet. The Claude pass runs 14 days for $700 of quota with $380 per week, priced at $50.97 for 14 days, and its quota is drawn at the duty-pass rate x2.5, worth at most about $280 of usage at list prices, less on models whose pass rate sits above list. The GPT pass runs 30 days for $3000 of quota with $900 per week, priced at $81.53 for 30 days, with the same x2.5 draw and at most about $1200 of usage at list prices, less on models whose pass rate sits above list. Other models always use the wallet, so a pass is an addition to the setup rather than a replacement for it.

That is the first hour in full: sign up, tag and bind your keys, match the client to the lane, trace one request in the usage log, know how to retire a leaked key, and understand how the wallet and the duty pass settle. Create your Relay365 account now, and let the desk take the watch from there.