Generation is the easy half. What makes this usable is that the same server takes the project to a downloadable MP4 and then to a public link, and passing a previous chat id back into a generation call keeps refining that project instead of starting a new one.
An MCP server for Tellers AI video creation. A prompt creates or edits a video, local footage can be uploaded as source material, projects render to a downloadable MP4, and assets publish to a shareable link — with generation sessions kept so one project can be refined across several conversations.
- A video created or edited from a natural-language prompt (tellers_generateVideo)
- Local files uploaded to use as source footage (tellers_uploadMedia)
- A project rendered to a downloadable MP4 (tellers_exportProject)
- An asset published with a shareable preview link (tellers_makeAssetPublic)
- Recent assets and exports browsed (tellers_listRecentAssets)
- Past generation sessions listed and their message history read (tellers_listChats, tellers_getChatMessages)
- Available AI models and account configuration inspected (tellers_getSettings)
Node.js 18+ and a build of the repo, plus a Tellers account and an API key created from the user menu at app.tellers.ai. Operations consume Tellers credits; new accounts start with some.
