What the listing writer produces
The output is structured, not a blob of prose. Every run comes back split into a product title, a bullet list, and a product description, with a notes section for anything the model wants to flag. That structure exists because marketplaces ask for those fields separately, and copy written as one long paragraph then chopped into a template always reads badly.
The inputs are equally deliberate. Alongside the raw product details you give a brief, a marketplace and positioning note, a buyer profile, and a tone. The buyer profile is the field that changes the output most: copy aimed at a specialist who already knows the category reads nothing like copy aimed at someone comparing options for the first time.
Five task presets sit on top of the same engine for the jobs people actually repeat, so you are not rebuilding a prompt every time you list something. Pick a preset, paste the details, and adjust the tone if the default does not match your brand voice.
- Product title, bullet points, and full description in every run
- Buyer profile and marketplace shape the register and the emphasis
- Presets for the repeated listing jobs
Where your text goes, stated plainly
This tool does not run in your browser. The fields you fill in, including the brief, the product details, the context, the audience, and any extra instructions, are sent over HTTPS to a Cleanor server, which forwards a prompt to a language model provider and sends the answer back. There is no local mode and no offline option, and any claim that nothing leaves your device would be false for this page.
That has consequences worth taking seriously before you paste. Do not put unreleased pricing, supplier terms, customer records, or anything under NDA into the form. Treat it the way you would treat any third-party writing service: fine for product marketing copy, wrong for confidential material.
Runs are recorded. The inputs you submit and the generated output are stored against your account so you can find a previous run again, and a separate metering row records token counts and cost. If you would rather a particular run not be kept, do not submit it.
Sign-in, credits, and the caps that exist
Because each run costs real money to serve, the workflow requires a signed-in account on a plan that includes monthly AI credits. A run deducts credits from that allowance, and if the balance is too low the request is refused with a clear message rather than silently producing nothing.
There are also rate limits and site-wide caps. Individual accounts are limited to a short burst of runs in a five minute window and a larger number per hour, which stops a runaway loop from draining an allowance. Above that, the whole site has a daily ceiling on both the number of AI runs and the amount spent serving them, so at busy times a run can be turned away until the next day.
This is a deliberate design choice rather than an accident. A tools site that leaves an expensive endpoint uncapped eventually gets one, from someone scripting it. The caps keep the feature affordable enough to stay free of per-run charges for the people using it normally.
- Requires sign-in and a plan with AI credits
- Per-account burst and hourly limits protect your allowance
- A site-wide daily run and spend cap can defer a request to tomorrow