Media Lab

A staging area for images, video and audio you want to work with — process them with an AI model, or hand them to a coding agent. Add one from the tile picker as Media Lab.

Adding media

Drop files onto the tile, or click it to choose files. Each file can be up to 100 MB, and a workspace holds up to 1,000 assets or 1 GB across all its Media Labs. Each asset previews in place, with playback for audio and video.

Per asset you can Send to coding agent, Download to your machine or Delete. Select several to send them to an agent together, download them as a zip, or delete them.

Processing with AI

Select an asset, pick a prompt, and run it:

• Presets cover the common jobs, filtered to what fits the asset: Bug report w/ timestamps, Describe in detail, Transcribe, Extract on-screen text, UI → component spec, Summarize + action items and Generate alt text. You can also write your own prompt.
• Model — choose a model that can handle the asset: Gemini works with images, video and audio; Claude, OpenAI and Grok work with images. A model marked Connect key needs your own API key first. You can also send the asset and prompt to a coding agent instead.

The result appears under Output, Rendered or Raw. Copy it, Send output to agent, or clip it into a Stash.

AI processing needs a Pro plan or higher: 50 runs a day on Pro, 500 on Enterprise. Sending media to a coding agent isn’t limited.

Sending to a coding agent

Choosing Send to coding agent copies the file into the agent’s sandbox and stages a prompt pointing at it for you to review and send. With one agent open it goes there directly; with several you pick; with none you can send it to a new quick lane. Terminal tiles can receive files too.