Models provide the base intelligence available to agents. Conceptually, an agent is a model plus whatever custom capabilities (tools, prompts, etc.) that you build. The Ardent platform has support for models from the three providers (OpenAI, Anthropic, and Google), and we’re planning to add support for open source (or open weights) models (Llama, Mistral, DeepSeek, etc.) in the near future.

Assigning a model

The Ardent SDK provides adapters for each of the model providers supported by the platform. For example, this agent will use Anthropic’s Claude Opus 4 model:
import { Agent, anthropic } from '@ardent-ai/sdk';

const agent = new Agent({
  model: anthropic('claude-4-0-opus')
});

export default agent;
You can read more about the adapters and the settings available for each model in the SDK reference:

Model API keys

While Ardent is in early access, you will need to provide your own API key to allow your agents to call out to the model provider of your choosing.
1

Choose a model provider

Select the model provider that you’d like to use. Each model has strengths and weaknesses, as well as different costs, so you might need to experiment to find the one that works best for your use case.(If you aren’t sure, we recommend starting with Anthropic!)
2

Create a developer account

Create a developer account with your chosen model provider.
3

Get an API key for your account

Find or create an API key using the model provider’s web app, and copy it to your clipboard.
4

Create a secret for your API key

Navigate to the secrets page in your workspace settings in the Ardent app and create a new secret. We recommend using the default name for the secret, but you can change it if you choose.
5

Add the secret name to your agent code (optional)

If you used the default secret name in the previous step, you won’t need to do anything special in your agent code. If you chose a custom name, you’ll need to reference it in your agent. For more information on how to set a custom secret name, see the SDK reference.

See also