How Claude Sonnet 5.5 API Proxy Refunds Work on SwiftRelay

First lapsEN2026-11-04

Your SwiftRelay pre-flight checklist before the first top-up

Getting started with SwiftRelay should feel like a clean first lap rather than a long detour through settings and billing menus. The service is a pay-per-token API relay: you top up a balance, connect, and each call deducts only what it used. There is no monthly fee and nothing to subscribe to, so you can start small, watch the usage list, and decide what to do next from real numbers.

Before you run your first request, work through this short checklist. First, sign up on SwiftRelay and note the sign-up gift credit of $0.22 sitting in your account. Second, choose a top-up amount that suits the way you plan to experiment. Third, connect your client to the API host https://api.fastapirelay.com with your swft_ key. Fourth, run a call and open the usage list, where tokens and cost per call are shown. That is the whole rhythm: top up, connect, run, read the usage, and repeat.

SwiftRelay is built for people who want the connection to be simple. The promise is straightforward: connected in minutes, copy three lines and run. No monthly commitment is waiting for you at the finish line, and no unused seat is being billed while your project sits idle.

Choosing a top-up that matches your working style

The available top-up amounts are $1.39, $6.94, $13.89, $27.78, $69.44 and $138.89. Pick the one that gives you room to test without turning the exercise into a financial decision. If you are simply checking how the setup behaves, a smaller amount keeps the whole thing light. If you expect a busy day of calls, a larger balance means fewer top-up stops in the middle of your workflow.

Because SwiftRelay bills per token, the balance moves with your usage rather than with a calendar. Each call deducts only what it used, and the usage list shows tokens and cost per call so you can follow the pattern yourself. For a quick sense of the pricing page, Claude Sonnet 5.5 is listed at $0.9792 per 1M input tokens, $4.89 per 1M output tokens and $0.0861 per 1M cache read tokens.

Example only: suppose a day of your work used 100,000 input tokens and 10,000 output tokens on Claude Sonnet 5.5. At the listed prices, the input would be $0.09792 and the output would be $0.489, for a total of $0.58692 for that day on that model. The point of the example is not a forecast for you, but a demonstration of how a modest day of usage maps directly onto the balance.

Your own mix will depend on the models you call. GPT-6.1 Sol is listed at $0.1736 per 1M input tokens and $0.8542 per 1M output tokens, while DeepSeek V4 Flash is listed at $0.2639 per 1M input tokens and $1.06 per 1M output tokens. Gemini 3.1 Pro is listed at $1.35 per 1M input tokens and $7.81 per 1M output tokens. With figures like these visible, the balance becomes a working instrument instead of a mystery number.

How a SwiftRelay refund works, step by step

Here is the refund procedure in plain terms. SwiftRelay refunds unused credit within 11 days of a top-up, minus what was used and the payment fee. In other words, the window is tied to the top-up itself, and the refund reflects the credit you genuinely did not consume.

Step one is to look at the 11-day window for the top-up in question. Step two is to check what has already been deducted, since used credit is removed from the refundable amount. Step three is to allow for the payment fee, which is also taken off. Step four is to contact the team through Messages in the dashboard, where support is handled and a login is required. There is no support email to hunt for; the dashboard is the channel.

Two details deserve emphasis. First, the refund is calculated on unused credit, so it is most useful when a top-up is larger than the work you actually ran. Second, the timeline starts from the top-up, so it is sensible to check the window early rather than after it has passed. If you top up, test lightly, and find that most of the balance is still sitting there, the refund route is there for you.

Keep your own record as you go. The usage list shows tokens and cost per call, and the refund subtracts what was used, so a clear view of your calls makes the arithmetic easier to follow. A simple habit is to check the usage list after each session and note the top-up date next to it. That is the sort of small discipline that keeps a stopwatch honest.

Connecting in minutes and running your first call

Once your balance is ready, the connection step is short. SwiftRelay uses the API host https://api.fastapirelay.com, and your swft_ key is the credential you paste into your client configuration. The setup is designed around three lines that you copy and run, so the barrier between signing up and seeing a response is deliberately low.

There is one important pairing rule to keep in mind: each model family works with its own client and protocol. Claude models answer only through the official Claude Code or Claude Desktop client, version 2.1.285 or later. Gemini models use only their native protocol, through Gemini CLI, and are not OpenAI-compatible. Codex, Cursor, Cline, Roo Code, Cherry Studio, OpenCode, SDKs and cURL use the OpenAI-compatible endpoint and pair only with GPT, DeepSeek and the other remaining models, never with Claude or Gemini.

For example, the Claude Code configuration is:

01  export ANTHROPIC_BASE_URL=https://api.fastapirelay.com
02  export ANTHROPIC_AUTH_TOKEN=<your swft_ key>

claude

The Codex configuration is:

01  export OPENAI_BASE_URL=https://api.fastapirelay.com/v1
02  export OPENAI_API_KEY=<your swft_ key>

codex

And the Gemini CLI configuration is:

01  export GOOGLE_GEMINI_BASE_URL=https://api.fastapirelay.com
02  export GEMINI_API_KEY=<your swft_ key>

gemini

For OpenAI-compatible clients, use the base URL https://api.fastapirelay.com/v1 with a swft_ key as the API key. After that, run one short call, open the usage list, and confirm that tokens and cost per call appear. That first clean run is the moment SwiftRelay stops being a configuration task and starts being a working tool.

Keeping the account tidy as usage grows

SwiftRelay supports up to 20 live API keys per account, which gives you room to separate projects, clients or experiments without piling everything into one credential. As you add keys, keep a note of which key belongs to which project so that a later audit of usage is quick rather than archaeological.

Support is available through Messages in the dashboard, with a login required, and there is no support email. That means questions about a top-up, a refund window or a specific call belong in one place, alongside the usage data that explains them.

Referrals are another part of the balance picture. You receive 14% of each top-up by someone you invited, and 3% from the people they invite. Those rewards can be withdrawn, with a minimum of $4.86, or moved into the balance to fund future calls. If you intend to use the balance for more work, moving rewards in keeps the accounting simple; if you prefer cashing out, the $4.86 withdrawal minimum sets the threshold.

The overall shape is what makes SwiftRelay easy to value: no monthly fee, nothing to subscribe to, per-token deduction, visible usage, a defined refund route and a small sign-up gift of $0.22 to begin. You are not committing to a plan; you are topping up a balance and paying for what you actually use.