A $0.90 Cursor API Receipt, Line by Line

A $0.90 API receipt is easy to explain: $0.30 for input and $0.60 for output. Whether your Cursor session costs that amount depends on the model, its actual rates, and how many tokens Cursor sends. Start with a transparent worksheet, then replace its assumptions with your account’s recorded usage.

This walkthrough connects Cursor to CheapRelay through an OpenAI-compatible endpoint. The receipt below uses illustrative rates, not a CheapRelay price quote or a measured Cursor benchmark. Any Cursor subscription charge belongs on a separate line.

01 / The $0.90 receipt uses two explicit assumptions

For this example, assume input costs $0.30 per million tokens and output costs $3.00 per million tokens. Assume the recorded usage contains 1,000,000 uncached input tokens and 200,000 output tokens, with no separately billed cache activity.

• ■ Uncached input: 1,000,000 ÷ 1,000,000 × $0.30 = $0.30.
• ■ Output: 200,000 ÷ 1,000,000 × $3.00 = $0.60.
• ■ Example API total: $0.90.
• ■ Cursor subscription: separate, if applicable.

Those token counts describe an accounting example, not a promised number of prompts or completed coding tasks. A short instruction can trigger substantial input when the editor includes files, conversation history, or tool results.

02 / 1. Create a key and copy the actual model ID

Sign up for CheapRelay, create an API key, and check the current rates for your intended model before sending requests. CheapRelay bills per real token, including cache reads and writes, with no monthly fee or subscription.

If you choose Gemini 2.5 Flash, copy its exact API model identifier from the service’s model listing. A display name and an API identifier are not necessarily interchangeable. Also record the input, output, cache-read, and cache-write rates that apply to your selection.

Keep the key private. Store it in Cursor’s credential settings rather than a project file that could enter version control.

03 / 2. Set Cursor’s base URL to the /v1 endpoint

Open Cursor Settings and find the model or API-key configuration area. Labels and placement can change between versions. Look for the OpenAI API key field and the option to override the OpenAI base URL.

• ■ API key: paste your CheapRelay key.
• ■ OpenAI base URL: enter https://api.cheapapirelay.com/v1.
• ■ Model: add or select the exact supported model ID you copied.
• ■ Custom configuration: enable the relevant key and base-URL overrides.

The service’s root is https://api.cheapapirelay.com, but this setup uses its OpenAI-compatible routes under /v1. Enter the base URL only; do not append /chat/completions. The separate /v1beta route serves Gemini’s native API and is not the route for this configuration.

04 / 3. Verify one small request before using a repository

Run Cursor’s verification action if your version provides one. Then open a fresh chat, select the configured model, and ask it to reply with a single word. Avoid attaching files for this first check.

Confirm that the request appears in your relay usage records. A successful response alone does not establish which billing path handled it. Check the model, timestamp, token usage, and charge where those details are available.

Cursor’s custom-key support depends on the feature and version. Do not assume that configuring chat also redirects autocomplete, background work, or every agent feature. Check support for the particular workflow you intend to use.

05 / 4. Replace the sample receipt with recorded charges

After a small coding task, calculate each billing category separately: tokens divided by 1,000,000, multiplied by the applicable per-million rate. Add uncached input, output, cache reads, and cache writes as your usage records define them. Avoid counting cached tokens twice.

If authentication fails, check the key. If the model is rejected, check its exact identifier. If requests reach an unexpected route, check the base URL and duplicate /v1 segments.

To control spending, begin with a narrow file selection and request a focused change. Then inspect the receipt before expanding the task. The useful number is the recorded cost of your workflow; the $0.90 example gives you the arithmetic to audit it.