Command line power, natural language ease
Download
Tell it what you want to happen, and it takes action:
Substage converts natural language prompts into command line commands.
Just tell Substage what you want to do with the files you have selected.
Make a jpg plz
Using an AI model such as GPT-4o, it creates a command to be run on your Mac’s terminal.
sips -s format jpeg ocean.png —out ocean.jpg
Substage evaluates the risk of running the command and might require confirmation to proceed. You can sanity check the command if you'd like.
Any output is boiled down into a summary of what happened.
Substage is free to try for 2 weeks.
$0.00
$7.99
$79.99
Substage is designed to help you work with files and folders on your Mac more efficiently by translating natural language into terminal commands. To do this, we send your prompt, along with some contextual information, to AI providers such as OpenAI.
One key thing to be aware of is that, by default, Substage also sends the paths of the files and folders you’re working with. This helps the AI make smarter decisions—like suggesting a zip file should be named “screenshots.zip” if you’re selecting multiple files named “screenshot1.jpg” and “screenshot2.jpg.” However, we understand that even sending file names is a privacy concern, and we’re exploring ways to limit this. In the future, we’d like to provide an option where only file extensions (like .jpg or .txt) are shared, while actual names remain local.
Substage does not aim to access or send the content of your files. However, there is a potential exception: when summarizing the output of commands, some file content could appear in the output. For example, if you run a command that prints a file’s contents to the terminal, and you ask Substage to summarize that output, the AI provider will process the contents of the file. We don’t process or store this data ourselves, but it’s something to keep in mind.
We've carefully selected our AI providers based on their privacy policies. OpenAI, Anthropic, and Mistral all have similar privacy commitments: they retain data for only 30 days and solely for abuse detection purposes, not for training their models. We've chosen not to integrate with Google Gemini because their privacy policy is not quite as clear.
We'd love to support offline AI models in the future so that your data never has to leave your machine—but we haven't explored this yet. Our goal is to strike the right balance between utility and privacy, and we'll aim to keep improving as we go.
We'd love to hear your feedback or suggestions on how to improve this further!
Looking for ideas? Here are some more example commands to try.
These make use of the bundled ffmpeg and ffprobe tools.
Anything else I should add to this list? Let me know!
