Set up your own model

You connect your own model on the Model page of the Captain app. The dashboard panel has no model settings, but it uses the model once it is connected.

Check the Requirements first.

Connect your model

1. Open Model in the Captain app. The page is titled Your own model.
2. Under Connect your model, fill in Endpoint. Select OpenAI, Anthropic or Gemini to fill it in for those providers, or type your provider’s address.
3. Enter the Model name, exactly as your provider spells it.
4. Check Context window (tokens) and change it to match your model.
5. Paste your API key.
6. Turn on Reads images only if the model accepts images.
7. Select Check and save.
8. The page asks Send your content to this provider? and names the address your messages and profile content will go to. Select Confirm. Changing any field cancels the question.

Captain then makes a test call to your endpoint. If it passes, you see Your model is ready and Captain uses your model by default. If it fails, That didn’t work tells you why, and nothing is saved.

Your key cannot be read back

The API key is write-only. Captain encrypts it, uses it only for your Captain messages, and never shows it again. The page shows only the last four characters, so you can tell which key is stored.

• To change the model or other fields on the same endpoint, leave the key field empty. Captain keeps the stored key.
• To change the endpoint, enter the key again. A stored key is never sent to a different address.
• To replace the key, paste the new one and select Check and save.

Test it again

The box at the top of the page shows Using followed by your model’s name, when it was last checked and the end of the key.

Select Check again to repeat the test call, for example after you changed something at your provider. If a message to your model fails later, the box changes to Your model had a problem with the provider’s error.

Note

You can run up to 10 checks in 10 minutes. After that you see “Too many model checks. Try again in a few minutes.”

Choose who answers in a chat

With a model connected, the message box shows a switch:

• Your model: your model answers and no credits are used. The message box says which model is answering.
• Credits: the built-in model answers and the message uses Captain credits.

Your model is the default. Your choice is remembered per profile for the browser tab you are in. You cannot switch while a reply is running.

Switch back to the built-in model

For one chat, set the switch to Credits.

To stop using your model everywhere:

1. Open Model.
2. Select Remove my model.
3. In Remove your model?, select Remove.

Your key is deleted, and Captain goes back to using your credits.

If your plan changes

If your plan no longer includes Bring Your Own Model, the saved model stays unused and Captain answers with credits. The Model page shows Not on your plan. You can still delete the stored key with Remove my saved model.