AI Assistance

Bring Your Own Key

Introduction

By default, Assistants run on the free QuantConnect Meta Agent, which requires no setup. Organizations with a paid Assistant Node can instead supply their own API key and run the Assistants on an LLM provider of their choice. Bring your own key when you want to run a specific model, tune how the model reasons, or route the cost through your own provider account.

The provider, API key, and model are stored in a secret. To configure your own provider, either add an assistant or update an existing one. In the Provider section of the Assistant Configuration, click the Use Your Own Provider option and select a secret from the Provider Token dropdown. The following settings become available:

SettingDescription
Reasoning EffortThe amount of reasoning the model applies before it responds. The default is high.
Max Turns The maximum number of turns the assistant takes per deployment. The default is 100.
Tool ChoiceThe tool selection policy. The default is auto, which lets the model decide when to call tools.

Secrets

A secret securely stores an LLM provider token and the model configuration that goes with it, so you can reuse the same credentials across your Assistants without pasting the token each time. To manage your secrets, log in to the Algorithm Lab and, in the left navigation bar, click Organization > Secrets. The Secrets page lists every secret in your organization with its name, key, configuration, and a masked preview of its token. The token is write-only, while the secret configuration stays visible and editable.

Add a Secret

Follow these steps to add a secret to your organization:

  1. On the Secrets page, click New Secret.
  2. Fill out the following fields:
    FieldDescription
    NameA label to identify the secret.
    KeyThe type of secret. Select Llm Provider Token to store LLM provider credentials.
    ProviderThe LLM provider that issued the token.
    TokenThe API key from your provider account.
    ModelThe model to run. The list shows the models that the selected provider offers.
  3. Click Save.

Update or Delete a Secret

On the Secrets page, each secret has Update and Delete actions. Click Update to edit the secret configuration. Because the token and its configuration are saved together, leave the Token field blank to keep the current value, or re-enter the token to change any configuration below it. Click Delete to remove the secret.

Use a Secret

To run an Assistant on a secret, associate the secret with the Assistant in one of the following ways:

  • When you add or update an Assistant, in the Provider section, click Use Your Own Provider and select the secret from the Provider Token dropdown.
  • When you deploy an Assistant, click Configure, select Bring your own token, and select the secret from the Secret dropdown.

In either dropdown, select Create new secret to add a secret without leaving the Assistant.

Keep the following in mind when you use a secret:

  • When you run an Assistant Team, the orchestrator's secret is used for the whole team.
  • You cannot change the secret after you deploy a task. To use a different secret, start a new task.

Claude

To run the Assistants on Anthropic's Claude models, follow these steps to get an API key:

  1. Log in or create an account on the Claude Console.
  2. On the API Keys page, click Create Key.
  3. Copy the key and paste it into the Token field in the Organization's Secrets.

OpenAI

To run the Assistants on OpenAI's models, follow these steps to get an API key:

  1. Log in or create an account on the OpenAI Platform.
  2. On the API keys page, click Create new secret key.
  3. Copy the key and paste it into the Token field in the Organization's Secrets.

OpenRouter

To run the Assistants on a model from OpenRouter, follow these steps to get an API key:

  1. Log in or create an account on OpenRouter.
  2. On the Keys page, click Create Key.
  3. Copy the key and paste it into the Token field in the Organization's Secrets.

You can also see our Videos. You can also get in touch with us via Discord.

Did you find this page helpful?

Contribute to the documentation: