How a Monthly Pass Quota Is Used, Request by Request

A monthly pass can make API spending easier to plan, but its headline allowance does not necessarily tell you how many requests you can send. The useful question is: what does each request deduct? This walkthrough explains the accounting step by step. The numbers below are hypothetical, not published Relay365 prices or pass terms.

#1 · Identify what your quota measures

Start with the unit shown in your pass details. An allowance might be expressed as money, credits, tokens, or requests. These are different accounting systems: a credit balance does not automatically translate into a fixed number of conversations.

Also check which models and endpoints qualify, when the allowance resets, and whether unused quota expires. Do not assume that “monthly” means the first day of each calendar month; confirm the actual billing period.

#2 · Separate the pass from your cash balance

A pass allowance and a pay-as-you-go balance may be separate funding sources. Before sending traffic, establish which source pays for an eligible request and what happens when the pass cannot cover it.

Relay365 is a pay-as-you-go relay desk for multiple model families. If you are using a monthly pass alongside it, check the current pass terms rather than assuming every request uses the allowance first. Model eligibility, fallback billing, and exhaustion behavior need explicit confirmation.

#3 · Follow one request from submission to usage

Suppose you send a request using Claude Haiku 4.5. Your client supplies the prompt, conversation history, and any other supported inputs. The service processes the request and produces a response.

In a token-priced arrangement, both input and output can contribute to the charge. A short question with a long answer can therefore cost more than another short question with a brief answer. Request count alone is not a reliable spending meter.

Check the applicable rate information for any additional categories, such as cached input or tool usage. Only include categories that the service actually bills.

#4 · Calculate the deduction using the pass rules

For a simple hypothetical example, imagine a pass containing 100 credits. Assume an eligible request has a final assessed cost of 0.8 credits, with no other adjustments. Its deduction leaves 99.2 credits.

Now imagine a second request costs 1.3 credits. The remaining allowance becomes 97.9 credits. The basic ledger is:

New balance = previous balance − eligible request deduction.

If the pass uses model-specific conversion factors, discounts, or separate allowance pools, apply those rules before calculating the remaining balance. Do not assume a displayed cash price equals the pass deduction.

#5 · Repeat the calculation for each request

Your third request does not inherit the second request’s cost. It gets its own usage assessment. Changing the prompt length, response length, or model can change the deduction.

For example, switching from Claude Haiku 4.5 to Gemini 2.5 Flash means checking the applicable rate and pass eligibility again. Longer conversations can also send more history on later turns, depending on how your client constructs requests. One visible chat session can therefore contain many separately accounted API calls.

#6 · Reconcile the ledger, not just the headline balance

After a small test, compare the usage records with the allowance movement. Look for the model, usage quantities, assessed charge, and funding source wherever those fields are available.

• Check whether several requests ran concurrently.
• Look for automatic retries or background calls from your client.
• Distinguish a final charge from any temporary reservation, if the service uses reservations.
• Confirm how failed, interrupted, and streamed requests are treated.

These are rules to verify, not universal billing behaviors.

#7 · Plan for the last request before exhaustion

Finally, find out what happens when the remaining quota is insufficient: rejection, another funding source, or another documented outcome. Configure application-side spending controls where available rather than relying on an assumed cutoff.

Relay365 is worth considering when you want a relay that works with Claude Code, Codex, Gemini CLI, and OpenAI SDK clients. Its API base is https://api.api365relay.com; OpenAI clients add /v1. Create a Relay365 account, review the current allowance and billing terms, and begin with a small test. The goal is a repeatable ledger: each request has an explainable deduction, and each deduction leads to a balance you can reconcile.