Attachments & Voice Input

Attachments. You can give the agent files — images, audio, video, PDFs, text — using the paperclip button on the chat pill, or by dragging-and-dropping or pasting onto it. Uploads are stored per-session and survive restarts, and show up inline in the chat (image previews, audio/video players, download links). The agent reads them with a built-in tool when relevant.

Images when the agent's model can't see. If the agent's model is text-only, there are two paths: a vision model can describe the image in words so the description folds into your message, or an agent with the Image Vision ability can actively ask a vision model specific questions about it. Either way you can attach an image and get useful answers even from a text-only agent.

Voice input. The microphone button is dictation: it uses your browser's speech recognition to transcribe what you say straight into the text box, where you can edit before sending. (The mic is hidden on browsers that don't support it.)

Related: Chatting with an Agent, Image Vision (Reading Images).