Filing issues breaks your flow. You context-switch out of the code to write up something you just saw, and detail gets lost on the way.
How it works
Record what you saw
Capture the bug, the idea, or the change directly from your screen.
Your agent reads it
Paste the Give to your agent prompt into Claude Code. It reads your words, your marks and an image of each mark.
Issue, fix, proof
It files one GitHub issue anchored to files and lines, works it, and can film the result.
- Stay in flow: record instead of writing tickets
- Issues anchored to the files and lines your asks point at
- Proof of the fix comes back as a narrated video
A day of screen recording for developers who work with a coding agent
Connect once
Run claude mcp add --transport http --scope user vidmatic https://api.vidmatic.ai/mcp in Claude Code. Then run /mcp, choose vidmatic and choose Authenticate. Your browser opens, you sign in, and the tools appear in the same session with no restart and no key to copy. Cursor works too; the MCP server page has its steps. To remove it later, run claude mcp remove vidmatic -s user.
Record instead of typing
When something looks wrong, record it. The Chrome extension lets you circle the broken element while you talk. New Recording in the app records any screen or window, which suits a desktop app or a terminal. Your agent gets your words with their timestamps either way.
Hand it over
Choose Give to your agent and paste the prompt. The agent reads the transcript, every mark and the image of each mark, searches your repo and files one GitHub issue, with a row per ask and one to three file and line anchors for each. Vidmatic never reads your code. Your agent reads it on your machine, and you can watch each stage on Agent Workstreams while it works.
If a teammate already filed an issue from a recording, run /mcp__vidmatic__issue_intel with the issue URL. Your agent posts every ask from the recording as one comment, with its file and line anchors.
Get proof back
When the change is done, /mcp__vidmatic__record films your running app with Playwright and narrates it. For a bug, /mcp__vidmatic__before_after records the path before the fix and again after it, and Vidmatic plays the two back to back. Both land in your Inbox.
Questions
- Does Vidmatic read my code?
- No. Your coding agent reads your code on your machine. Vidmatic gives the agent the recording: the timed transcript, your marks and an image of each mark.
- Which coding agents does it work with?
- Claude Code and Cursor. The MCP server page lists the steps for each.
- Do I need an API key?
- No. You sign in from the browser when Claude Code asks you to authenticate. If you prefer a key, Settings → Developer → Build my command gives you the same line with your key in it.
- What if the transcript is not ready yet?
- The agent waits the number of seconds the response gives and asks again. A processing transcript does not mean the recording is silent.
- Can my agent read a recording a teammate sent me?
- Yes. Open the share link and use Give this to your agent. The prompt it copies includes --token, which lets your agent read that one recording.
Related use cases
- Screen Recording for AI Coding Agents
Screen recording for AI coding agents: record your screen, circle it and say it. Claude Code or Cursor reads the transcript, marks and frames over MCP.
- Bug Reports From Video
Make a bug report from video: record the bug and talk it through. With GitHub connected and auto-create on, Vidmatic opens the issue; or hand it to your agent.
- Generate a Demo Video
Finish a feature in Claude Code and let it record, narrate and publish the demo. Vidmatic drives your running app, adds a voice-over and returns a share link.
Show Your Agent the Screen
Connect once, record once, and hand your agent the recording.
Explore more Vidmatic use cases
Docs: Quickstart