How Margaret works
The whole idea is to write your instructions once, then let Margaret handle the task the same way every time.
The create → run → improve loop
1. Create a helper
Tell Margaret what you want help with in plain English. She turns it into a helper — a structured set of instructions, input fields, and an expected output format. You can start from scratch or install a ready-made bundle.
2. Give it an input
Paste in text, fill a short form, or point to a local file. Every time you run a helper, you give it the same kind of input — your messy notes, a client brief, a document to summarise. Margaret knows what to do with it.
3. Get your output
Margaret runs the helper using your AI provider and saves the result as a readable text file on your computer. You can copy it, open it, or use it however you like. Nothing is sent anywhere you didn't expect.
4. Improve the helper
If the output isn't quite right, ask Margaret what's vague, missing, or unclear in the helper's instructions. She'll suggest specific edits. Apply the ones that make sense, then run it again. The helper gets better each time.
Three things that make it different
A prompt is something you write fresh each time. A helper is something you build once and reuse. Margaret stores the instructions, the input format, and the expected output so you never have to explain the task again.
Your helper library, your run history, and your outputs are all plain files on your Mac. Margaret never stores your work on our servers — there are no servers. The only external contact is your AI provider when you choose to run a helper.
Nothing runs automatically. You trigger every run yourself. Before each one, Margaret can show you exactly what will be sent to the AI — so you're never guessing what left your machine.