Skip to content

Learn how to launch Antigravity CLI, collaborate with an autonomous local agent, review generated files, and execute terminal test commands.

This guide walks you through a rapid onboarding exercise. In this exercise, you direct an autonomous agent to create a Python utility script, review the changes, and verify its execution.

Follow these steps to create and verify a Python script with the agent:

  1. Create a clean project directory and launch the Antigravity TUI:

    mkdir agy-demo && cd agy-demo
    agy
  2. Prompt the agent to generate a Python scraping script:

    Type the following instruction in the prompt box at the bottom of your screen and press Enter:

    Write a simple python script to fetch web page text

    The agent reads the workspace, determines that no files exist, and formulates a plan to create a script. You receive real-time updates as the agent performs reasoning and schedules actions.

  3. Open the artifact review screen to inspect the proposed code:

    Once the agent finishes generating the file, a notification appears. Press Ctrl + R to enter the Artifact Review screen and complete the review:

    • Navigate to the newly created main.py using ↑/↓.
    • Review the complete file content and diff.
    • Press Y to approve the creation of main.py.
    • Press Esc to close the review panel and return to the primary prompt.
  4. Execute a test command with the agent to verify the output:

    Direct the agent to run the Python script to verify its behavior. Type the following command in the prompt box and press Enter:

    Run the python script and show me the output

    The agent proposes to run python3 main.py. Press Y to confirm and execute the command. The agent runs the script locally and streams the standard output directly into your terminal screen.

  5. Exit the Antigravity session:

    Once you complete your task, press Ctrl + D (or type /exit) in the prompt box to close the TUI and restore your original shell session.

Now that you’ve executed your first agent-assisted workflow, learn how to configure the CLI and master core concepts: