Here’s a deep dive on using the CLI for AI Dev.

Usage

In the most basic form, you can run the CLI using npx, and it’ll prompt you for information.

npx ai-dev@latest

See Getting Started for information on this.

Non-Interactive Usage

However, what if you’ve decided to write a script or CI runner that calls Magnet Agent?

That’s what our non-interactive CLI is for:

OPENAI_API_KEY=your_openai_api_key_here npx ai-dev@latest -f ./ -t "Your task here." -o ./output

Tasks

It’s possible to pass tasks to the CLI in two formats:

• Inline, with -t. Useful if the task is short.
• As a path to a file with the task, with -tf. Better if the task is fairly long.

Parameters

• f is the path to the folder you want to run the task on.
• t is the task description.
• tf is a file that contains a task description.
• o is the path to the folder where the agent will output the results.
• c will prompt the agent to interactively clarify the task description before implementing.
• a will cause the agent to try to apply the changes to the source directory.