Skip to content

What the agent can do ​

TIP

Haven't installed the skill yet? Run npx skills add https://ai.modelfold.online — see Install modelfold-models.

What modelfold-models lets your agent do, and how to ask for it.


Supported model types ​

TypeExample request
Image"Use GPT Image 2 to make a 1:1 product photo of a white ceramic mug, save it to ./public"
Video"Use Seedance to make a 5-second 720p clip of a paper boat drifting down a rainy street, save it to ./media"
Speech"Read this product intro aloud with a TTS model and save it as intro.mp3"
Chat"Use Claude on ModelFold to tighten this copy"

The model list is live. The agent reads it from ModelFold every time instead of relying on remembered names, so new models on the marketplace work right away.


What the agent does for you ​

  1. Finds the model. Reads the live model list and picks a model and price that fit. Name a model if you care which one.
  2. Calls the right endpoint. Uses the image, video, audio or chat endpoint for that model type, with parameters like aspect ratio, duration, resolution or voice.
  3. Waits for the result. The request stays open until generation finishes (videos usually take 1–5 minutes).
  4. Saves the output to the path you gave — and, if you ask, writes the integration code into your project.

How to ask ​

  • Say where to save, e.g. ./output or ./public/hero.png.
  • Name the model if it matters: "use GPT Image 2", "use Veo 3.1".
  • Give the key parameters: aspect ratio, resolution, duration, voice. Anything left out uses the model default.
  • Input images must be public URLs (image-to-image, image-to-video). Uploading local files isn't supported yet.
  • Ask for code when you want code: "add an endpoint to my API that calls this model on ModelFold".

Good to know ​

  • Every successful call uses credits, the same as calling the API yourself; failed jobs are refunded automatically. See usage and your ledger in the console.
  • Video is priced by duration × resolution and speech by text length — the agent will tell you the expected cost of expensive jobs first.
  • The agent reads your key from MODELFOLD_API_KEY and never writes it into the project.

ModelFold — one API for all the best AI models.