Set up your OpenAI API key
Connect CandidTalentEdge to OpenAI so you can set up a role and review candidates. You do not need to write code or use a terminal.
An API key is a private credential that lets the app use OpenAI on your behalf. OpenAI bills your API account for that work. This is separate from both your ChatGPT subscription and your CandidTalentEdge license, with no CandidTalentEdge usage markup.
Start setup · Fix a setup problem
Before you begin
- CandidTalentEdge installed on an Apple Silicon Mac. Download the app if needed.
- An internet connection and an OpenAI API account with permission to create a key and use API billing. For a work account, ask your administrator which OpenAI project to use.
- Permission to send the relevant job and candidate content to OpenAI. Records are stored locally, but AI processing is not offline. Read how candidate data is handled.
Connect in five steps
-
Set up API billing
Sign in to the OpenAI API platform, or create an account. Open API billing and follow the account’s instructions to add payment details or credits. Check that you are in the organization that should pay for this work.
A paid ChatGPT plan does not pay for this API usage. If billing settings are unavailable, ask the organization owner to complete this step.
-
Choose a project and spending controls
An OpenAI project groups keys and usage. Use your organization’s designated project, or create a dedicated CandidTalentEdge project in OpenAI settings to keep this usage separate.
In that project’s Limits, review Spend. Set alerts for an amount you want to track. Alerts notify you; they do not stop spending.
To stop requests at a monthly threshold, choose Edit spend limit, enter a Monthly spend limit, turn on Enforce a hard limit, and select Save. Enforcement can lag, so recorded spend can slightly exceed the amount. Organization limits can also stop requests. If you cannot manage these settings, ask your administrator. OpenAI explains spend controls.
-
Create a key for the app
Open OpenAI API keys in the intended project and use the key-creation control. Give the key a recognizable name, such as “CandidTalentEdge Mac”. Copy the secret key when it is shown and keep it private.
Use a project API key, not an administrative key. If your organization restricts key creation or permissions, ask its administrator for access to model discovery and model requests. You do not need to install any developer tools.
-
Save it in CandidTalentEdge
Open the desktop app and select Settings. Find OpenAI API key, paste your key into API key, then select Save.
The app shows API key saved and Configured. The entry field clears after saving; this is expected. The saved key is stored in macOS Keychain.
Configured does not mean connected. Continue to the connection test. You do not need to enter the key again simply because the field is empty.
-
Test the connection
In the same section, check AI model. Start with the available option marked (Recommended). Leave Reasoning effort at its app default for now.
Review the startup disclaimer if the app asks you to do so. When you are ready, select Test connection. This sends a small billable request to OpenAI. You can use Cancel while it runs.
Success: look for “Connected to OpenAI with” followed by the selected model’s name or identifier. “Connection not tested this session” is not a successful test. If the test fails, use the recovery steps below.
Next: set up your first role
Return to Your Jobs and select Create New Job. Have the job description ready and follow the role setup steps. Review suggested screening filters before using them; you can start without custom filters.
Begin with one candidate you are authorized to process. Check the findings against the source material before adding a larger pool. Role setup and candidate processing use the API and may incur charges.
Keep an eye on usage
CandidTalentEdge shows Recorded API cost for each job: an estimate based on recorded usage for candidate reviews and related work. Some requests may not return complete usage information, and some activity may not be captured, so the estimate may not include every charge.
For a fuller view, open OpenAI Usage and select the relevant project and date range. Use API billing for payment and credit information. If usage is not visible, ask your administrator to check your access.
Costs depend on the model and the work performed. Connection tests, repeated evaluations, and additional AI operations can add usage. Check your first small batch before processing more. See current OpenAI API prices.
If setup does not work
After addressing the problem, return to Settings and select Retry connection or Test again, whichever the app shows. Each new test may incur a charge.
The key is rejected
For “Invalid API key format” or “Authentication failed. Check the API key.”, make sure you pasted the complete secret key, not its name, a masked value, or your ChatGPT password. Save it again. If it has expired, been revoked, or is no longer available to copy, create a replacement in the correct OpenAI project.
A rate-limit or billing problem persists
The app may show “OpenAI rate limit reached. Try again later.” for both temporary request limits and billing-related limits. Pause before retrying. If it persists, check API billing, credit balance, and both project and organization spend limits. A ChatGPT subscription does not resolve an API billing problem.
Restore credits or ask your administrator to review the applicable limit. Do not raise a spending limit unless you intend to authorize more usage. Replacing a key alone does not fix exhausted credits.
The app cannot reach OpenAI or the test times out
Check your internet connection. On a managed network, ask IT whether access to OpenAI is allowed; do not bypass workplace security controls. Retry when connectivity is restored. A timeout does not necessarily mean the key is invalid.
The selected model is unavailable
In AI model, select Retry discovery if shown, then choose an available reviewed model. If no models are available or OpenAI rejects the test, ask your administrator to check the project’s model access and key permissions.
The key will not save, or Test connection is disabled
Read the message next to the control. Resolve any macOS Keychain access prompt, then restart the app if it asks you to. A disabled test may instead require a saved key, acceptance of the startup disclaimer, or attention to trial or Pro access. Replacing the key will not resolve those access requirements.
If the app says OPENAI_API_KEY overrides any key saved in Keychain., a launch-time setting is taking precedence. Ask whoever configured it to remove that override and restart the app before testing the saved key.
Still stuck? Contact CandidTalentEdge support with your app version and the error message. Remove keys and candidate information from anything you share. Never send your API key to support.
Keep your key private
Treat the key like a password that can authorize spending. Do not put it in screenshots, email, chat, shared documents, or source code. If you keep a backup, use a password manager.
If the key is exposed: revoke it in OpenAI API keys promptly, review usage for unexpected activity, and create a replacement. Paste the replacement into CandidTalentEdge, select Save, and test it. Other tools using the revoked key will also need updating.
Clear in CandidTalentEdge only removes the saved Keychain entry. It does not revoke the key at OpenAI or undo charges. For planned replacement before expiry, save and test the new key before revoking the old one.
Sources and verification
Checked on against the app’s Settings and connection-test implementation and official OpenAI documentation. Account screens and permissions can differ. Key creation, billing changes, and a paid connection test were not performed for this guide.