Screen recording for AI agents: add the Vidmatic MCP in one line, hand off a recording

This is screen recording for AI agents, with Vidmatic. Your coding agent can't watch a video, but it can read one: a timed transcript, timed marks, and pictures of what you circled. Record two ways. The Chrome extension lets you draw and talk. The New Recording button needs no extension; your voice carries the asks. Setting up takes one line. In Settings, under Developer, one click builds the command, with your key already inside. Prefer no key? Paste this one line instead, and Claude Code adds Vidmatic, waiting for you to sign in. Open the M C P menu, choose Authenticate, and sign in once in your browser. That's it. Vidmatic shows connected, and its tools are ready in this same session, with no restart. Now the hand-off. This recording of a pricing page has three marks: a circled button, a crossed-out banner, a wrong price. On the recording's Agent tab, Give this to your agent holds a single line. Click Copy prompt. Paste it into Claude Code. The agent reads every spoken line with its time, and each mark lined up with what you said. Then it opens each mark as an image, sees exactly what you circled, and searches your code for the words on screen. And it writes back issue intelligence: one row per ask, in your words, with the time, the mark, and the file and line. Record it, paste one line, and your agent knows exactly what you asked for.