Questions,
answered
What makes the images, what Logospell adds, what a set costs, how long you have to fetch one, and what changes when your agent can ask for art itself.
How the images
are made
What generates the images?
Frontier third-party image models, chosen for quality and swapped as better ones ship. Logospell trains none of its own.
What does Logospell add?
Everything around the call.
- One style and one subject list come back as a coherent set, not a pile of unrelated pictures
- Every image sized, centered and named after its subject, on the solid or transparent background you asked for
- Changed for free afterwards: size, margin, format, and the background too on a transparent set
- The same set exports as icon trees for the web, iOS, Android and Flutter
Why not call an image model myself?
Because of what a set is: 18 icons drawn by one hand, at one scale, on one background, each named after its subject and sized for where it is going. Ask a model 18 times and you get 18 strangers, then the work of making them match. Logospell takes one style and one list and hands back the set, in one call.
What happens
to your sets
How long do my downloads last?
30 days from the moment a set is generated. Your generations page shows the deadline on every set and hands the download link back if you lose one. The window never extends, so save the files when they arrive.
Can I change a set after it is generated?
Yes, free and as often as you like: size, canvas, minimum margin, sizing, format and quality, plus the background on a transparent set. The same download link serves the new delivery. Style, subjects and references belong to generation rather than delivery, so a different picture is a new set.
What does it cost?
A set is 1 credit, whether it holds one subject or 18. A transparent set is 1 credit, a single illustration 1 credit. Editing a set and exporting it as icons cost nothing. New accounts start with 3 free credits, no card, and credits do not expire.
What an agent
can do with it
Do I need an AI agent to use this?
No. The generate page makes sets in the browser: sign in, describe a style, list your subjects. MCP is for when an agent is doing the building and would rather fetch its own art.
Can it make a single picture instead of a set?
Yes, over MCP. generate_illustration composes one picture from one prompt: a scene, a banner, a hero image, a portrait. You choose its shape and size from a fixed list of pixel dimensions, and it comes back as PNG, JPEG or WebP at the quality you name, ready for your agent to build with.
What does connecting an agent unlock?
Your agent asks for the art itself, in the middle of building, and gets files back it can use straight away. It can carry a style across later calls to grow a set, change how a set is delivered with edit_image_set, and export icons for every platform with export_icons, without you leaving the conversation to go find images. The home page has one build done twice, with the same code both times and the art the only difference. Connecting a client takes a key and one config block.