Get File

This tool finds files in your library by name, your own description, or the text parsed out of them — so a PDF is findable by a phrase on page four even when its filename says nothing. Results are metadata (name, type, folder, parse status); the agent reads a file with Get File. You can narrow to images, PDFs, or everything else, or to a single folder.

This tool uses image generation models from Google or OpenAI to create images from text descriptions. Supports long descriptions. Choose a size (s, m, l, xl) and an aspect ratio (square, landscape, or portrait); on OpenAI models, small and medium always render as 1024x1024 (so pick large or extra large for non-square output) and large square is also capped at 1024x1024. Generates one to four images per request.

This tool uses OpenAI GPT image models to edit images the user has attached. Common uses include making a background transparent, changing the style while keeping the composition, combining multiple reference images into one, or applying targeted edits. Requires one or more images attached to the current message.