Claude Code API relay pricing: what Relay365 actually speaks
Which protocols the Relay365 desk actually answers
A careful buyer usually starts with a simple question: which protocols does the relay endpoint speak, and can my existing client use them without a rewrite? The Relay365 desk answers that question with a short set of lanes and firm client rules, because every lane sits on one table and every request can be traced in the usage log.
Relay365 is an independent third-party relay with a single API base. The model wire name is the routing key. A Claude lane uses a wire name such as claude-opus-5, a GPT lane uses a name such as gpt-5.6-sol, and a Gemini lane uses a name such as gemini-2.5-flash. The pricing page shows the full table, so this article stays with the protocol question instead of reciting every lane and every rate.
Here is the practical answer. Claude models work only from the official Claude Code and Claude Desktop clients, on Claude Code 2.1.285 or newer. Gemini runs on its native protocol only, through Gemini CLI. The remaining lanes pair with Cursor, Cherry Studio, Cline, OpenCode, SDKs, and cURL-style tooling, including the GPT, DeepSeek, Kimi, GLM, Qwen, Grok, MiniMax, and Hunyuan lanes.
Relay365 keeps the wiring visible in the dashboard. API keys carry tags, a key can be bound to a lane, quick-wiring snippets sit next to the key list, and the interface is available in six languages. That is how a buyer checks the protocol fit before paying, rather than after.
Claude Code and Claude Desktop: one client rule, no exceptions
Claude Code is the tool most people search for first, so the Relay365 desk is plain about it. Claude models work only from the official Claude Code and Claude Desktop clients, with Claude Code 2.1.285 or newer. They do not work from Cursor, Cherry Studio, Cline, OpenCode, any SDK, or cURL. If your team standardises on one of those other clients, the compatible lanes are GPT, DeepSeek, Kimi, GLM, Qwen, Grok, MiniMax, and Hunyuan.
Once you are inside Claude Code or Claude Desktop, the request travels to the Relay365 endpoint, and the wire name decides which lane draws from your balance. Relay365 prices four token types separately: input, output, cache write, and cache read. That separation matters, because a request that recalls cached context is not billed the same way as a fresh input.
For example, if you send 1 million input tokens and 1 million output tokens on Claude Opus 5, the Relay365 rates are $2.02 for input and $10.07 for output, so the example cost is $12.09. The Relay365 output rate is about 40% of the official list price of $25 for the same token count. The pricing page shows the full table for every lane.
The duty desk also keeps the ledger honest. Every call is settled on the token counts the upstream reports, and a call that fails upstream with a 502 or 503 is not billed. A disconnected stream is billed for what was actually used, which keeps the accounting tied to real consumption rather than intent.
Gemini CLI: native protocol, two endpoints, no chat completions
Gemini CLI is the second tool worth naming, because Gemini does not follow the same path as the other lanes. Gemini uses its native protocol only, through Gemini CLI. It cannot go through /v1/chat/completions, so a request shaped for an OpenAI-compatible endpoint will not find a home here.
The native surface is compact: :generateContent for a single reply, and :streamGenerateContent?alt=sse for streams. That is the whole Gemini protocol story at the Relay365 desk, and it is why a Gemini lane is worth checking against your client before you commit any spend.
Pricing follows the same four-token breakdown. For example, if you send 1 million input tokens and 1 million output tokens on Gemini 2.5 Flash, the Relay365 rates are $0.2389 for input and $1.91 for output, so the example cost is $2.1489. No official list price is quoted for this family, so there is nothing to compare it against, and Relay365 does not pretend otherwise.
Cursor and other clients: pick a compatible lane
Cursor is the tool many teams search for when they want a pay-as-you-go setup without a monthly fee. On the Relay365 side, Cursor, Cherry Studio, Cline, OpenCode, SDKs, and cURL-style tooling pair with the GPT, DeepSeek, Kimi, GLM, Qwen, Grok, MiniMax, and Hunyuan lanes.
That separation is the point of the desk. Rather than one blurred endpoint, Relay365 keeps the lanes distinct, tags each API key to a lane, and shows quick-wiring snippets in the dashboard. You can see the wiring before you commit spend, and every request can be traced in the usage log.
The usage log shows the latest 200 requests with the time, model, input, output, cache recall, charge, and result of each call. It does not show which key made the call, so treat it as an account-level trail rather than a per-key audit. The ledger alongside it tracks the wallet, bounty, and pass accounts, and in-site messages reach the desk once you log in.
Request handling has stated guardrails. HTTP 429 is returned when there are more than 120 calls in 60 seconds on one key, or more than 8 requests in flight for one account. Knowing these conditions in advance makes the setup easier to plan, because the desk explains the rule instead of leaving you to guess.
Two duty passes for steady traffic
For Claude and GPT, Relay365 offers optional duty passes. Other models always use the wallet. A pass comes with a dedicated key that never touches the wallet, which keeps pass traffic separate from prepaid balance traffic in the ledger.
The Claude 30-day pass at $3900 of quota costs $506.53 for 30 days and has no weekly ceiling. The GPT 30-day pass at $2350 of quota costs $81.53 for 30 days, with at most $700 of quota per 7 days. In both cases, quota is drawn at the duty-pass rate x2.5, which works out to at most about $1560 of usage at list prices for the Claude pass and about $940 for the GPT pass, and less on models whose pass rate sits above list.
One behaviour is worth reading twice. 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.
Otherwise, billing is straightforward: a prepaid wallet, no monthly fee, and every call settled on the token counts the upstream reports. A new account also receives $0.25 in signup credit, and the referral bounty pays 18% at level 1 and 6% at level 2 of what referred members pay, withdrawable from $11.11 or movable into the wallet. Refunds are available within 14 days of a top-up for unused balance, minus what was consumed and the payment channel's fees, while bounty and promotional credit are not refundable.
Bring your questions to the Relay365 desk with your client and your traffic pattern in hand. The desk will confirm the lane, the protocol, and the numbers before you commit, and you can create a Relay365 account and read the full pricing table on the site.