Skip to main content

Custom Models

With your own API key (BYOK, "bring your own key"), you can use models from your own provider account in ConsoleX chats, tasks and agents. Calls to these models are billed by your provider, not from your ConsoleX plan allowance or credits. Bring your own key is included in every plan.

Supported providers​

Add Custom Model currently offers these providers:

  • OpenAI
  • Anthropic
  • Microsoft Azure (see Azure OpenAI)
  • OpenAI Compatible: any endpoint compatible with the OpenAI API
  • Google Gemini
  • Mistral
  • Ollama (see Ollama)
  • Amazon Bedrock
  • Groq AI
  • Qwen
  • OpenRouter
  • Cohere
  • Google Vertex (see Google Vertex AI)
  • xAI
  • DeepSeek

Add a custom model​

  1. Click … (More) in the left rail, then Models. If the Models do the work intro appears, click Skip to the model list.
  2. In the Model Hub, click Add Custom Model at the top right.
  3. In Select Model Provider, choose your provider.
  4. Fill in the connection details the provider needs, such as your API key, the model name and, for some providers, a base URL or region. Then save.

The Select Model Provider dialog opened from Add Custom Model in the Model Hub

Your model now appears in the Model Hub under Private. The Model Hub can also test a model's connection in one click, which is a quick way to confirm your key and endpoint work. Your key is stored encrypted.

Use your model​

Pick your model wherever you choose a model:

  • In a chat, from the model menu in the composer.
  • In a task, under Done by → Model in the task wizard or on the task's Config tab.
  • In an agent, under More settings → Which model to use → Choose a model.

To make it your default model, click Set default next to it in the Model Hub. See Use shared models.

How BYOK usage is billed​

  • Model calls made with your key are billed by your provider. They use no ConsoleX plan allowance or credits.
  • Checks and ConsoleX's built-in tools still use your plan's usage. For example, the automatic check after a task run uses the checking model set in Settings → Checking model, and a web search during the run uses the built-in search tool.

See Credits and usage for how plan allowance and credits work.