Questions,
answered
What it makes, what it costs, who owns the results, and what your agent can do with it.
What Logospell
makes for you
What is Logospell?
It turns a few lines you write into a matching set of images, every image its own file: an image set, a character set or an illustration set. On the generate page, or from your AI agent.
What is an image set?
A style and a list of subjects. The style says how everything is drawn, in words, in up to 3 reference images, or both; each subject is one image, such as an icon, a sprite, a logo or a piece of an interface. Each is drawn on its own, on one solid color or a transparent background. Up to 16 subjects at a time.
What is a character set?
A character and a list of variations. The character is who or what appears in every image, the same mascot, creature, person or product, in words, in up to 3 reference images, or both; each variation is what it is doing in one image. Each is drawn on its own, on one solid color or a transparent background. Up to 16 variations at a time.
What is an illustration set?
A style and a list of scenes. The style says how every illustration is drawn, in words, in up to 3 reference images, or both; each scene is one full composed illustration, such as a setting, a banner, a hero image or a poster. Illustrations made together share one size, chosen from a list. Up to 6 scenes at a time.
What makes the images?
Frontier third-party image models, chosen for quality and speed, and swapped as better ones ship. Logospell trains none of its own.
Why not ask an image model myself?
Ask a model for images one at a time and each drifts from the last: line weight, palette, lighting, framing, and for a character, the character itself. Logospell is built for sets, so a whole set arrives in one style and a character keeps its look across every variation. Each image comes sized for where it goes and named after what it shows, ready to drop in or to export for the web and apps at every density. And a set can grow later without drifting.
Is the transparency real?
Yes. A set delivered transparent arrives as PNG or WebP with a real alpha channel, not a painted checkerboard, so every image drops onto any background.
Can the style come from my own art?
Yes. Add up to 3 images of your own, alone or with a written style. The set takes on how they are drawn, not what they show.
Can I add to a set later?
Yes, while the set is still downloadable: more subjects, variations or scenes, drawn as the set was, on the generate page or through your agent.
Working with
your sets
What sizes and formats can I get?
An image or character set comes at its native resolution, which is how to get larger images, or at any width and height you fix from 256 to 512 pixels, with the margin you choose, and an illustration set at one of a fixed list of sizes. Each comes as PNG, WebP or JPEG, and a set delivered transparent as PNG or WebP. WebP and JPEG take the quality you set.
Can I change a set after it is made?
How it is delivered, yes, free and as often as you like: its size, margin, format and quality, and on a character set or a transparent image set its background (any color, or transparent). An illustration set keeps its shape, changing to its full size or an exact half, quarter and so on of it, and its format and quality. What the pictures show cannot change: a different style, subject, variation, scene or character is a new set.
Does changing a set lose quality?
No. Every size, format and export is cut from the full-resolution original the model produced, never from an earlier delivery, so a set changed ten times is as sharp as the first.
Can a set become app icons?
Yes, free: one zip with drop-in icon trees for the web, iOS, Android and Flutter at every density, from the generations page or through your agent. An illustration set exports the same way, its illustrations kept whole for an app or site.
How long can I download a set?
30 days from when it is made. Adding to a set restarts the window for the whole set, as does using one of its images as a reference for another generation. The generations page shows each set's deadline and hands its link back. Nothing else extends it, so save the files when they arrive.
Do you keep my images?
Only for the download window: files leave our servers when it closes, and you can delete a set sooner on the generations page. Reference images you upload expire a while after their last use, and your agent can delete them sooner.
What a set
costs you
What does it cost?
Every set is 1 credit: an image set or a character set from one image to 16, an illustration set from one scene to 6. A pack of 10 credits is $10, so a set comes to $1. Editing and exporting are free.
How do I try it?
New accounts start with 3 credits, no card needed.
Do credits expire?
No. There is no subscription either: packs are one-time purchases, and their credits wait until you use them.
Am I charged if a generation fails?
No. A credit is spent only when a generation completes.
Can I get a refund?
Yes, for a pack you have not spent any of: email us within 14 days of buying it and it is refunded in full. The terms have the details.
Who owns
what you make
Who owns the images I make?
You do. Logospell claims no ownership and assigns you any rights it holds. AI images carry caveats, though: someone else may get a similar one, and copyright may not protect it, so review an image before relying on it commercially.
Are my prompts or images used for training?
No. Logospell does not train models on your prompts, references or images, and does not sell your data. The privacy policy lists who handles what.
Isn't this trained on artists' work?
Logospell trains no models of its own. It generates with commercial image models, and what those were trained on is decided by the companies that build them.
What an agent
can do with it
Do I need an AI agent?
No. The generate page does it all in the browser. MCP is for when an agent is doing the building and should fetch its own art.
Which agents work with it?
Anything that can send your API key with each request: most coding agents, IDEs and agent SDKs, and the Claude apps on plans whose connectors offer request headers. Connect has setup ready for Claude Code, Claude, Cursor, VS Code, Devin Desktop, Cline, Antigravity, Gemini CLI and Grok Build, with your key filled in.
Why MCP instead of an API?
Your agent knows what images the build needs, so it asks mid-build and uses the files straight away. MCP lets an agent do that with no glue code.
Can it make a single picture?
Yes: an illustration set of one scene. An illustration set composes up to 6 scenes, banners, hero images or posters in one style, every illustration made together at one size from a fixed list, as PNG, JPEG or WebP, on the site or over MCP.
Can my agent buy credits?
It can ask for a checkout link, which costs nothing. You complete the payment yourself, and the credits land on that key within seconds.
Is it open source?
The hosted server is not; it is the product. The plugins, setup docs and examples in the public repo are MIT.