$1 Test Budget: Connect Cursor to a Custom API

01 / Set a $1 test target before opening Cursor

Use $1 as a personal spending target for your first Cursor connection test. That is a suggested experiment budget, not a quoted deposit requirement, model price, or guaranteed spending cap. The goal is simple: get one short coding answer through your own API endpoint, check the charge, and decide whether to continue.

Cursor charges and API charges are separate. Adding your own key does not automatically remove a Cursor subscription requirement or make every Cursor feature available through that key. Check your current Cursor plan before paying for anything else.

02 / 1. Get your relay key and exact model ID

Sign up for CheapRelay, then create an API key in your account. CheapRelay offers pay-as-you-go access with no monthly fee or subscription. Billing follows actual token usage, including cache reads and writes.

Before adding funds, check the available payment amounts and current model pricing. If the minimum payment exceeds your experiment budget, decide whether you want that much prepaid credit. A balance is money committed even when you have not spent it on requests.

• ■ Copy your API key and keep it private.
• ■ Find the exact API model ID in the relay’s current model listing.
• ■ Record the input, output, and applicable cache token rates.
• ■ Check whether the account provides a spending limit, balance alert, or usage dashboard.

Gemini 2.5 Flash is one option to inspect. Choose based on its current listed price and compatibility with your intended Cursor workflow. A display name is not necessarily the model ID you must enter.

03 / 2. Enter the OpenAI-compatible base URL in Cursor

Open Cursor Settings and find the Models area. Depending on your version, API controls may appear under Models, API Keys, or a similarly named section. Look for the OpenAI API key field and the option to override the OpenAI base URL.

• ■ Paste your relay key into the OpenAI API key field.
• ■ Enable the custom base URL override.
• ■ Enter https://api.cheapapirelay.com/v1 as the base URL.
• ■ Save or enable the configuration using the controls shown.

The /v1 suffix matters: it selects the OpenAI-compatible API routes. Do not paste a complete request endpoint such as /v1/chat/completions into a field that asks for a base URL. CheapRelay also has Gemini-native routes under /v1beta; those are not the routes for this setup.

04 / 3. Add the model and select it explicitly

If Cursor offers an Add Model control, enter the exact model ID from the relay. Enable it, open a chat, and select that model explicitly. Avoid an automatic model selection mode during this first test so you can more easily confirm which model handled the request.

An OpenAI-compatible endpoint does not guarantee support for every Cursor feature. Custom-key support can differ between ordinary chat, agent tools, and other features. Start with a plain chat request; test file editing or tool use separately after the basic connection works.

05 / 4. Send one tiny request and inspect the receipt

Open an empty folder or a small scratch project. Send this prompt: “Write a JavaScript function that adds two numbers. Return only the function.” Keep large files, terminal logs, and repository context out of the request.

After the answer arrives, inspect the relay’s usage record. Confirm the model, token counts, and charge where available. Usage reporting may lag, so avoid sending repeated requests while waiting. A successful answer alone does not establish that the intended billing route was used.

Estimate later costs as input tokens multiplied by the input rate, plus output tokens multiplied by the output rate, using the provider’s stated units. Account for cache reads and writes at their listed rates. One short test cannot predict the cost of an agent session with repeated calls.

06 / 5. Fix connection errors before adding more credit

• ■ Authentication error: check the key and account status.
• ■ Model not found: compare the entered model ID with the current listing.
• ■ Route error: check the base URL, including /v1.
• ■ Chat works but tools fail: check feature compatibility before retrying.

If available, set an account spending limit near your $1 target. Otherwise, monitor usage and stop early; manual checks are not a hard cap. Expand to one small coding task only after the first charge makes sense.