Why Read Usage Records Call by Call?
When you first sign up for an API relay, the honest question is simple: what did that one request actually cost me? SwiftRelay answers that in the dashboard, where every call is listed with its tokens and its cost. There is no monthly fee and nothing to subscribe to, so the ledger is the bill, and each row is a small, checkable fact rather than a rounded-up estimate you have to trust.
The API host for requests is https://api.fastapirelay.com. Once you top up a balance and copy your key into your client, calls start deducting from that balance, and the usage list shows what each one used. This article walks through what the records mean, how to check the arithmetic yourself, and what a first-time reader should look for before running a busy workload.
What Each Usage Record Shows
Each entry in the usage list reports the tokens used by a call and the cost of that call. The figure comes from the per-1M-token rates shown on the SwiftRelay pricing page, so it reflects the model you picked and the amount of work the request involved. You do not need to convert a single call into a monthly estimate to understand it, because the record already carries the cost for that request.
Think of the list as a lap-by-lap log for your own API traffic. A short prompt with a brief reply will sit near the top with a small cost. A long document, a heavy context payload, or a verbose answer will sit lower with a larger cost. Reading a few entries in order makes the pattern obvious, and the list makes it easy to spot which calls are expensive before you repeat them.
For example, suppose a single call uses 1,000 input tokens and 500 output tokens on a model priced at $1.53 per 1M input tokens and $7.27 per 1M output tokens. The input portion is $0.00153 and the output portion is $0.003635, giving a total of $0.005165 for that request. This example uses the published Claude Opus 5.5 rates, and the same arithmetic applies to the other models on the pricing page with their own input and output prices.
How to Check the Token and Cost Figures
Start with the model name on each row, because the rates change with the model. Then check the tokens and the cost side by side. If a call costs more than you expected, the record tells you whether the input was unusually large or whether the output ran long. Cache read entries, where applicable, are listed separately on the pricing page, so you can see how cached input is accounted for.
Here is the example calculation written out in full. For 1,000 input tokens at $1.53 per 1M tokens, the cost is $0.00153. For 500 output tokens at $7.27 per 1M tokens, the cost is $0.003635. The two figures add up to $0.005165. Because the usage list shows the tokens and the cost per call, you can confirm that the numbers in the dashboard match this kind of calculation instead of taking them on faith.
It also helps to remember what you are buying. SwiftRelay is a pay-per-token API relay: you top up a balance, and each call deducts only what it used. If you are used to paying a fixed subscription for an official account, a per-call view gives you a different kind of control, since every request can be checked on its own and nothing runs in the background on a recurring plan.
Copy Three Lines and Run Your First Call
Getting started with SwiftRelay is designed to be quick. You can create up to 20 live API keys per account, which is enough to keep a working key for everyday work and another for testing without juggling too many credentials. Each key draws on the same balance, and the usage records still break the traffic down call by call.
For an OpenAI-compatible client, the base URL is https://api.fastapirelay.com/v1 with a swft_ key as the API key. The three lines below are the whole setup:
01 export OPENAI_BASE_URL=https://api.fastapirelay.com/v1 02 export OPENAI_API_KEY=<your swft_ key>
codex
Once those lines are in place, run a small test request and then open the usage list. The first entry will show you the tokens and the cost, which gives you a clean baseline for the models you plan to use regularly.
What to Look for Before You Scale Up
Reviewing a handful of calls is the fastest way to build a habit around cost. Look for the expensive rows, note what made them large, and adjust the input or trim the requested output length. Over time, the usage records become a practical guide for your own prompts rather than an abstract bill.
Support is available through Messages in the dashboard, so if a row looks unclear you can ask there with the details in view. Login is required for that area, and there is no support email to hunt for. A fresh sign-up also includes $0.22 in gift credit, which is enough to make a first test call and see how the records look in practice.
Top-Ups, Rewards and Unused Credit
Because SwiftRelay sells no plans, your balance is yours to fill in the amounts you want: $1.39, $6.94, $13.89, $27.78, $69.44 or $138.89. Start with a modest top-up, run a few calls, and let the usage records tell you how fast the balance moves before you commit to a larger figure.
If you try SwiftRelay and change your mind, unused credit can be refunded within 11 days of a top-up, minus what was used and the payment fee. That keeps the first test honest: the money you did not spend is still recoverable under those terms.
There is also a referral reward worth knowing about: you receive 14% of each top-up by someone you invited, and 3% from the people they invite in turn. Those rewards can be withdrawn, with a minimum of $4.86, or moved straight into your balance to cover the next set of calls you want to read line by line.
No monthly fee, nothing to subscribe to, and a record for every call: sign up for SwiftRelay today, run your first request, and read the usage records yourself.