Lampo: keep the frame and drawing attached to your agent's revision task

[vll711]

We're building Lampo for the review step after an agent makes a video. HyperFrames' HTML/CSS workflow is a useful fit: the agent can change the source, while the person directs what should happen on screen.

Take a layered scene: “move this title” leaves plenty of room for interpretation. In Lampo, draw around the title on the exact frame and write:

Move this above the foreground layer. Keep the camera motion and entrance timing. I'll check the overlap in the next Version.

The agent reads the Note and marked frame through MCP, works on your source and uploads a new Version. You compare it and check the fix yourself. Lampo reviews exported video; there is no automatic HyperFrames integration being claimed here.

Try it with one MP4 from a real task: Lampo Cloud signup. Connect your agent in Settings → Connect an agent, then tell it Use Lampo for "<project>".

Source and local setup are open. For a local Codex setup, lampo mcp config codex prints the connection config.

For your next revision, does a marked frame capture enough intent, or do you also need a reference clip to explain the motion?