How to give Cursor video context

Here's how to give Cursor video context, in three steps. Cursor's agent reads text and images, not video files. So the trick is to turn your recording into a timed transcript and pictures it can fetch. Step one: add this single entry to your Cursor MCP file. There's no key. Cursor starts a browser sign-in, and you choose what the agent may do. Step two: record the bug with the Chrome extension. Say what's wrong and circle it. Each mark is saved with the sentence you were saying. Step three: in Cursor's agent chat, ask it to read the recording, look at each mark, and fix what you asked for. It gets the transcript with exact times, every mark with your words, and an image of each mark, so it sees the very button you circled. A few limits today: desktop Cursor only, skills run through get skill, and there are no console logs. Vidmatic keeps what you see, say and draw. The full guide, with setup, troubleshooting and an FAQ, is on the Vidmatic blog.