This panel runs on whichever AI provider you choose, using your own API key. Add one key to start. Add a second and every answer is marked twice, side by side, so you can see where two markers disagree.
Pick a question area and a severity, then choose whether to speak or type. The panel reads each question aloud. In speaking mode your answer is transcribed as you talk, so you practise saying it rather than writing it. Answers are marked on legal accuracy, structure, evidence and risk handling. Speech recognition needs Chrome or Edge; questions are read aloud in any modern browser.