New · Connected AI assistants

Tell your assistant. The project is born here.

“Instagram post, Saturday happy hour, for the pizzeria”: one sentence to ChatGPT or Claude, and a real project appears in Immaginia — 1080×1350 format, palette, copy and composition, as separate elements you then refine by hand in the editor.

Editable elements, not a flat imageFreePermissions you choose, revocable
1

Say what you need

One sentence is enough: the occasion, the tone, who it's for. No sizes, no technical terms.

2

The project is created in Immaginia

Your assistant builds a complete project: the right social format, a palette, headlines and copy already composed.

3

Open it and refine

Every text, photo and shape is a separate element: in the editor you move, change and fix whatever you want.

Two ways to connect it

One for people who just want it to work, one for people who work with a configuration file. They are not the same.

ChatGPT: it just connects

No key to copy and no file to edit: ChatGPT sends you to an Immaginia page, you authorize, and the connection is done. Three permissions are granted, spelled out on the consent screen:

The three permissions granted

  • read your projects (and only yours)
  • create new ones, complete projects included
  • upload the images the project needs

Claude Desktop, Cursor, Codex: personal key

Clients with a configuration file use a personal key instead of the authorized connection: you choose which permissions it carries, it expires on its own after 90 days, and you can revoke it at any time. Setup, the thirteen tools and the permissions are in the technical guide:

The full technical guide (docs/MCP.md)

What the assistant cannot do

This is not a list of promises: these are boundaries built in, and they hold for any assistant you connect.

It doesn't read other people's data

It sees your projects and only yours. The lists it goes through never contain email addresses in the clear.

It doesn't answer your customers for you

On support messages it can only prepare a draft: sending doesn't exist as an operation, and the final word stays with a person.

It doesn't suspend or promote anyone

On accounts and moderation it can only propose: every proposal waits until a person approves it from the panel.

It doesn't delete accounts

No permission allows it, even if you wanted to grant one. And every permission you do grant can be revoked at any time.

The questions worth asking

Do I need to know how to code?

No. From ChatGPT you add Immaginia and authorize the connection on one of our pages: no key to copy, no file to edit. The personal key is only for developer tools like Claude Desktop, Cursor or Codex.

Is it free?

Yes: connecting and creating projects costs nothing. There is a common-sense ceiling — up to 20 complete projects per hour — that normal use never meets.

What does the assistant see of my data?

Only what the granted permissions cover: your projects. It doesn't see other people's projects, it doesn't see email addresses, and the images in your projects are not poured into its replies.

Can it edit a project I already made?

From ChatGPT, no: the edit permission is not granted with that connection, so it can create, read and duplicate, but an existing project is yours to change in the editor. With the personal key used by developer tools, edits can be queued — and are confirmed by opening the project.

How do I disconnect it?

At any time: remove Immaginia from ChatGPT's connections, or revoke the personal key. From that moment the assistant can no longer get in, and every key expires on its own after 90 days anyway.

Your next project can start from a sentence.

Create your free account, connect your assistant, and open what you talked about in the editor.

Create your free account