kral Documentation
Sign in

Chat overview

The chat is where most of kral happens. One prompt box, every major model behind it, and a row of tools you can switch on per message.

The prompt box

  • Model picker (top left): choose Claude, GPT, Gemini, Grok, Mistral, or an open-source model. Switching mid-conversation keeps the full history. Details under Switching models.
  • Attach: drop PDFs, Office documents, images, or code into the box. See Files.
  • Tool badges: below the input you can toggle tools per conversation: web search, code interpreter, file search, and artifacts.
  • Microphone: dictate instead of typing. See Voice and read-aloud.

Under every answer

Each answer shows what it cost, split by input and output tokens. Copy, regenerate, branch off, or read it aloud from the same row. Regenerating or editing creates a new branch; nothing is lost. See Branching and forking.

The sidebar

Your conversation history, searchable. Bookmark conversations you return to, or run several models side by side in one view to compare answers.

Settings worth knowing

  • Presets: save a model plus parameters (temperature, system prompt) as a one-click preset.
  • Memory: opt-in personalization across conversations. See Memory.
  • Temporary chats: conversations that leave no trace in your history. See Temporary chats.

Next