Requirements

Bring Your Own Model lets Captain answer with an AI model from your own provider account instead of the built-in one. Messages answered by your model use no Captain credits; you pay your provider for the usage.

You need three things: a plan that includes the feature, an endpoint Captain can reach, and a model that can call tools.

A plan that includes Bring Your Own Model

The feature is part of some OceanResume plans. See the pricing page for which ones.

On a plan without it, the Model page in the Captain app shows Not on your plan with a See plans button.

An endpoint Captain can reach

The endpoint is the web address of your provider’s API. It must be:

• A public https:// address on the standard HTTPS port.
• A hostname, not an IP address.
• Free of extras: no user name or password in the address, and no query string. The API key goes in its own field.

Addresses on a private or local network are refused, so a model running only on your own computer does not work.

The endpoint must speak one of two APIs:

• https://api.anthropic.com uses Claude’s own API.
• Any other endpoint must speak the OpenAI chat-completions API. Enter the base address that /chat/completions hangs off.

The Model page has shortcut buttons that fill in the endpoint for OpenAI, Anthropic and Gemini. Any other provider with an OpenAI-compatible API works the same way: enter its base address yourself.

A model that supports streaming and tool calling

Captain changes your profile by calling tools, and it shows replies as they are written. The model you choose must support both:

• Tool calling (also called function calling).
• Streaming replies.

Captain tests this before it saves anything. It makes one small test call to your endpoint and asks the model to call a tool. The check fails, with a message that says what is missing, when:

• The model answers in text instead of calling the tool.
• The endpoint does not stream back a tool call.
• The provider rejects the key or does not know the model name.

A model that fails the check is not saved.

What else you enter

• Model: the model’s name, exactly as your provider spells it.
• Context window (tokens): how much text the model can take in at once. At least 4,000.
• API key: a key from your provider account.
• Reads images: turn this on only if the model accepts images. With it off, image attachments are not offered while your model answers.

Good to know before you start

• You have one own model for your whole account. It works in every profile’s chat.
• Your messages and profile content are sent to your provider, under that provider’s terms. See Limits and what is sent.
• There is a daily limit on messages answered by your own model.
• You can switch back to Captain credits at any time, per chat.

Ready? Go to Set it up.