YapSketch — Show Your Agent What You Mean
Description and use
YapSketch streamlines feedback to coding agents by combining voice and screen annotations into a single prompt, replacing complex multi‑app workflows.
Problem
Communicating detailed feedback and ideas to coding agents was cumbersome, requiring multiple apps and manual transcription.
Outcome
YapSketch fully replaced the author’s previous workflow (Wisprflow), enabling rapid, single‑step feedback delivery to coding agents and significantly reducing iterative exchanges.
Workflow
- 1. Press hotkey to start recording and drawing.
- 2. Speak feedback while drawing annotations on screen.
- 3. Stop recording; YapSketch transcribes speech locally and captures the annotated screen.
- 4. YapSketch pastes the combined text and image into the chosen coding agent’s prompt (Claude Code, Codex, Cursor, etc.).
Evidence
I made YapSketch to make it easier for me to communicate my feedback and ideas to coding agents.
Now I hit a hotkey, talk, and draw on my screen. When I stop, I paste it into Claude Code (or Codex, Cursor, whatever) and the agent gets what I said plus a picture of each thing I pointed at.
It fully replaced Wisprflow for me, and it cut down the back and forth a lot.
It's 100% free and open source, and speech is transcribed locally on your machine.
- Record date: 2026-10-20T20:41:47.302Z