The agent goes from a prompt to a narrated video and then keeps editing it by description — the step that normally pushes you back into an editor. Scene-level authoring sits underneath for the moments when describing the change is not precise enough.
VideoZero's own remote MCP server, at https://engine.videozero.ai/mcp. Connect an agent to it and making a video becomes something it does inside the conversation rather than a tab you switch to.
- Complete narrated videos generated from a prompt
- Revisions given in plain language against a video that already exists
- Direct scene authoring and editing, for the parts a description will not pin down
- Stock media and icon search, so the visuals get filled in without leaving the chat
- A read of how much of your usage is left
A VideoZero account. The first connection sends you through an OAuth sign-in; there is no key to paste. Claude takes custom connectors on every plan including Free, and on Team and Enterprise an Owner adds it in organization settings first. A ChatGPT custom connector needs a paid plan, or you can add the official VideoZero app instead.
One command — claude mcp add --transport http videozero https://engine.videozero.ai/mcp