Claude Code and Codex on Relay365 API Relay Keys
What a careful buyer asks before the first top-up
You are looking at Relay365 because you run Claude Code and Codex side by side, and you want to know what happens to your money before any of it moves. Fair questions. This is the duty desk of an AI API relay, run round-the-clock, and it is an independent third-party relay. Here is what we can tell you, plainly, with the numbers that sit on our own sheet.
First, where do the calls go and how are they billed? Relay365 keeps 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 at all. A stream that gets disconnected is billed only for what was actually used. That is the whole billing model, and it is the same whether you are wiring Claude Code or Codex.
Second, what does a token cost? Relay365 prices the four token types separately: input, output, cache write and cache read. On Claude Sonnet 5.5, that is input $1.18, output $5.91, cache write $1.48 and cache read $0.1028 per 1 million tokens. On GPT-6.1 Sol, it is input $0.1986, output $0.9944 and cache read $0.0389 per 1 million tokens. Separating those four rates is what lets a cache-heavy workflow cost what it should.
Why two keys, one per tool, one tag per project
The setup most teams land on is simple: one key for Claude Code, one key for Codex, and a tag on each key naming the project it belongs to. Relay365 gives you API keys with tags and a key bound to a lane, so the wiring is done at the key level rather than in a spreadsheet somewhere. A key bound to a lane means your Claude traffic and your GPT traffic never share a line, and a tag means you can tell at a glance which project a key was made for.
This matters when the bill arrives. Relay365 keeps a usage log of the latest 200 requests showing the time, model, input, output, cache recall, charge and result of each call. Every request can be traced there. Note one honest limit: the log does not show which key made the call, so the tag on the key is your own record of which project it serves. The dashboard also carries a ledger of wallet, bounty and pass accounts, quick-wiring snippets, in-site messages to the desk once you log in, and six interface languages.
On clients, the rules are firm and worth reading before you sign up. Claude models work only from the official Claude Code and Claude Desktop clients, Claude Code 2.1.285 or newer. Cursor, Cherry Studio, Cline, OpenCode, any SDK or cURL pair only with the GPT, DeepSeek, Kimi, GLM, Qwen, Grok, MiniMax and Hunyuan lanes. So your Codex key and your Claude Code key are genuinely different animals, and Relay365 keeps them that way.
What the rates look like next to official list prices
Relay365 quotes rates per 1 million tokens in approximate USD, with the account itself kept in CNY. Where the sheet carries an official list price, the comparison is worth stating. Claude Sonnet 5.5 lists at input $3 and output $15; the Relay365 output rate is about 39% of official. GPT-6 Astra lists at input $10 and output $50, and the Relay365 output rate is about 7% of official. Those are the only comparisons we make, because those are the only families where the sheet carries an official price.
Here is an example, worked only from those rates. Say your Claude Code session sends 2,000,000 input tokens and 500,000 output tokens in a day on Claude Sonnet 5.5. At Relay365 rates that is 2 x $1.18 plus 0.5 x $5.91, which comes to $2.36 + $2.955 = $5.315 for the day. At the official list prices it would be 2 x $3 plus 0.5 x $15, which is $6 + $7.5 = $13.50. That is the shape of the saving, and you can run the same arithmetic for your own mix.
For a second example, say Codex on GPT-6.1 Sol sends 1,000,000 input tokens and 200,000 output tokens in a day. At Relay365 rates that is 1 x $0.1986 plus 0.2 x $0.9944, which is $0.1986 + $0.19888 = $0.39748. Add cache reads at $0.0389 per 1 million tokens and you can see how the four separate rates add up line by line rather than being bundled into one number.
What a duty pass changes, and when the wallet is the answer
For Claude and GPT only, Relay365 offers duty passes, and they are optional. A pass comes with a dedicated key that never touches the wallet, which is a clean way to ring-fence a project budget. All other models always use the wallet, so if your project mixes in Kimi K3 or DeepSeek V4 Pro alongside your Claude and GPT work, the wallet is where that traffic settles.
Take the Claude 30 day pass at $1400 with a $450 per week cap. It costs $94.03 for 30 days and gives $1400 of official-price quota drawn at official price x2.5, which is about $560 of official-priced usage, with at most $450 of quota per 7 days. For a GPT-heavy project, the GPT 30 day pass at $3000 with a $900 per week cap costs $81.53 for 30 days and gives $3000 of official-price quota drawn at official price x2.5, which is about $1200 of official-priced usage, with at most $900 of quota per 7 days. The pricing page shows the full table of passes if neither of those fits.
Refunds sit on the wallet side. Within 14 days of a top-up, unused balance can be refunded minus what was consumed and the payment channel's fees; bounty and promotional credit are not refundable. A new account also starts with $0.25 of signup credit, which is a low-risk way to wire up both keys and watch the usage log fill in before you commit anything larger.
What holds steady when the traffic gets busy
Limits are stated up front so nothing surprises you mid-sprint. An HTTP 429 comes back when you exceed 120 calls in 60 seconds on one key, or more than 8 requests in flight for one account. Splitting Claude Code and Codex onto separate keys gives each its own call budget on that first limit, and keeping the tag on each key keeps the log readable when you are checking what happened.
That is the promise Relay365 makes in practice: every lane sits on one table, every request can be traced in the usage log, and the four token types are priced separately so you pay for the tokens you actually moved. Create a Relay365 account and wire up your Claude Code and Codex keys today.
Further desk bulletins
| Date | Bulletin title | Filed under | Language |
|---|---|---|---|
| 2026-10-29 | Shift log: Claude Code command time-savers, ranked | Agent shift notes | EN |
| 2026-10-29 | Duty Log: Keep the Contract, Clear the Agent Backlog | Agent shift notes | EN |
| 2026-10-29 | Shift Log: Shell Access Approved, Host Access Denied | Agent shift notes | EN |
| 2026-10-29 | Shift Log: Tests First Kept the AI Patch in Bounds | Agent shift notes | EN |
| 2026-10-29 | PR Review Shift Log: Agent Findings, Checklist Verdicts | Agent shift notes | EN |
| 2026-10-29 | Shift log: Claude Code team access routed and reviewed | Agent shift notes | EN |