Create an agent with Gemini

As an alternative to defining instructions manually for an agent, you can use the Start with AI feature when creating an agent. This is a quick way to use AI to generate an agent for you.

With this approach you provide:

  • A clear 1-2 sentence goal for your agent.
  • Optional knowledge documents.

Limitations

The following limitations apply:

  • This feature only works for the root agent. Multi-agent architecture is not yet supported.
  • This feature only works on empty agents.

Steps to generate an agent

To generate an agent:

  1. Using the agent builder, create an agent.
  2. Click the Start with AI button to open the Start with AI pane.
  3. Provide a clear, 1-2 sentence goal for your agent.
  4. Optionally add up to 5 knowledge documents (total size of all documents must be under 8MB) using the Upload button. These documents help teach your agent how to handle a variety of scenarios. You can upload files like tool catalog, requirements, transcripts, and FAQs.
  5. Click Build a preview.