Use your own AI key

Bring an OpenAI, Anthropic or OpenRouter key and AI limits no longer apply.

Before you start#

You need an account with at least one of these providers, with billing or credit set up, and an API key from its dashboard:

OpenAI: GPT models, billed by OpenAI.
Anthropic: Claude models, billed by Anthropic.
OpenRouter: one key for models from many companies, billed by OpenRouter.

You can save a key for all three. Only one is used at a time, and you can switch whenever you like.

Add a key#

Go to Settings, then AI provider.
On the provider's card, choose Add key and paste your API key.
Optionally pick a model. Leave it empty to use our recommended models for that provider.
Choose Test and save. We send the provider one tiny request to check the key works before anything is stored.

Settings, AI provider, then Add key on a provider's card

Paste your API key, and a model if you want one

We send the provider a tiny test request. Does it pass?

Yes

The key is encrypted, saved and turned on

Any other saved key turns off, but stays saved

Nothing is saved, and you see what went wrong

Fix the key or model and choose Test and save again

What happens when you add a key

Which key runs a request#

Every AI request, whether tailoring, an import, a draft or an inline edit, checks your keys first:

You tailor, import, draft or ask AI to edit

Is one of your keys turned on?

Yes

Runs on that key, with your chosen model or the provider's defaults

Billed by your provider. Plan AI limits don't apply.

Runs on Shortlist's own AI

Counts toward your plan's AI limits

How each AI request is routed

If the provider refuses a request, for example because the key ran out of credit, the request stops and tells you why. It doesn't quietly switch to your plan.

Pick a model#

The model box on each card is searchable and lists the models your key can use. You can also type any model ID. A model you choose is used for every AI step, and we test it with your key before saving it. Use Edit model to change it later; models you have used before stay in the list.

With the model left empty, these defaults are used:

Provider	Quick steps (imports, job posts, edits)	Tailoring
OpenAI	`gpt-5.4-mini`	`gpt-5.5`
Anthropic	`claude-haiku-4-5`	`claude-sonnet-5`
OpenRouter	`openai/gpt-5.4-mini`	`anthropic/claude-sonnet-5`

Small models can return changes that don't fit your resume. If a run comes back empty or off, try a stronger model.

Switch providers or go back to your plan#

Switch provider: turn on Use this key on another card. Your other keys turn off but stay saved.
Use your plan again: turn off the active key. Requests run on your plan and its limits until you turn a key back on.
Change a key: choose Replace key on its card. The new key is tested before the old one is replaced.
Delete a key: choose Remove key. If it was the active one, requests go back to your plan.

What changes on your own key#

AI requests run on your key and are billed by your provider.
Plan limits on AI tailoring, edits, imports and drafts stop applying.
AI runs in the same tab lists each request and its cost. Costs show as "Unknown" for models we don't have prices for; your provider's dashboard has the exact figure.

If something goes wrong#

You see	What to do
The provider rejected this key	Copy the whole key again, and check it hasn't been revoked in the provider's dashboard.
This key can't use that model	Check the model ID, pick one from the list, or leave the model empty to use the defaults.
Out of credit or rate limited	Add credit with the provider, or wait a minute and try again.
Your saved key can't be read anymore	Remove the key and add it again.
The model list doesn't load	Type the model ID directly; it is still tested.

How your keys are kept#

Keys are encrypted before they are stored, and only their last four characters are ever shown back. They are never included in exports or logs, and they are only sent to the provider they belong to. Remove them any time from the same page.