Claude Motion makes MP4s from code: when a Sume Agent run fits instead

Claude Motion is for animations built from your own content: text, charts, shapes and images moved around on screen and exported as MP4. A Sume Agent Completion is for a backend job that needs an agent to produce media and hand back URLs. If the video is a chart explainer for a meeting, Motion fits. If a service of yours must start the job and receive files, Sume's API is the surface that exists today.

What Claude Motion is, from coverage read 2026-10-09

Press coverage of the 2026-10-08 beta describes Motion as a way to turn text, diagrams and images into animated explainers that are editable as code and exportable as MP4. It is limited to Team and Enterprise plans. The coverage we read does not say which model or renderer sits underneath, so this page makes no claim about that.

Anthropic's own announcement post is linked below as a source. We could not read its full text, so every Motion fact on this page comes from the press summary. Treat plan limits as something to confirm on Anthropic's pages.

What a Sume Agent Completion returns

An Agent Completion runs the Sume Agent on one ad-hoc task. The Sume docs describe the same runtime as the Agents chat: sandbox, tools, MCP bridge and media generation. Your backend starts it with an API key and no person has to watch it.

The create call returns 202 and a receipt, not a finished video. You poll status_url, or register a run webhook, and a completed run puts generated media under output.images, output.videos, output.audio and output.files as media.sume.com URLs.

Choosing between them

The two differ on who starts the work and what is being drawn. Motion is a person-in-the-loop feature inside Claude. Sume's completion is an HTTP call that your code makes, with a spend cap you must set because nobody is there to approve spend.

Use Motion when the content already exists as numbers and slides and you want it animated. Use a Sume completion when the task needs new stills, clips, voice or music, or when the result has to land in your own system. For a fixed edit of assets you already have, the Timeline render is a cheaper, more predictable path than a full agent run.

• Chart or report explainer for a meeting: Motion.
• Generated clips or images returned to your backend: Agent Completion.
• Same task every week on a clock: a schedule.
• Same recipe with changing inputs: a Format run.

Before you wire either one in

Check the plan gate on the Claude side before you promise anything to a team that is not on Team or Enterprise. On the Sume side, create a new API key if yours predates Agent Completions, because older keys fail with 403 insufficient_scope and scopes cannot be added to an existing key.

If you want a webhook instead of polling, give a public HTTPS communication.webhook_url. Sume sends one signed POST when the run completes or fails, and none for a canceled run.