1. Enable billing Before proceeding to create an agent, you need to enable billing for the project. Billing is enabled by creating a usage-based Stripe subscription in Project Settings. Once you enable billing, your project will start incurring charges based on your usage. See Pricing for more details.
  2. Create an agent To create an AI agent, visit the Agents page in the admin dashboard and click the Create an agent button. Select the model that you want to use and name your agent.
  3. Configure the agent To update your agent’s configuration, visit the Agent Configuration page. You can set your own values for system message, initial turn, provider-specific options like temperature and frequency_penalty etc. When configuring the agent, you have access to dynamic API resources like user and agent. You can use the Handlebars template format to refer to these resources in your configuration:
    System Message
    You are an AI assistant that works for Acme Inc. Your name is {{agent.display_name}}. Your main task is to
help {{user.first_name}} with any questions they may have.
  4. Activate the agent Once you’ve configured your agent, you can then activate it so that it starts responding to user requests.
    Activation is a mechanism that allows you to enable or disable your agent. For example, if you want to immediately stop your project from incurring charges you can deactivate all your agents using the dashboard or the Admin API.
  5. All done! You can now test your agent in the Playground with a test user or start integrating it with your application using our official SDKs.