What the assistant does

One product is a few minutes in the app. A catalogue is the same few minutes, several hundred times, and that is the part worth handing over.

  • Import. import_product_from_url for a store page, or create_product_from_images when there is no page to read.
  • Wait for the draft. Poll get_product until the import and the creative direction have finished.
  • Read and steer it. The direction comes back as text — hook, content focus, format, restrictions. update_video_direction edits it, redraft_video_direction asks for another angle.
  • Render. generate_video. This is the only step that spends credits, and it needs your explicit go-ahead.
  • Collect. Poll get_video for the finished MP4 and a public share link.

Brands work the same way — list_brands, create_brand and set_product_brand — so a loop over a catalogue keeps every clip in one voice.

Why drive it this way

It asks before it guesses

Every video is drafted against a brand. An import whose storefront matches no brand you already have stops and asks rather than picking one. Guessing wrong is the failure you cannot afford when a script is running unattended.

Only one step costs anything

Reading a product page, drafting the direction and rewriting it are free. Rendering is the only tool that spends credits, and it will not fire without your say-so, so an agent cannot quietly run up a bill.

Nothing proprietary on the wire

JSON-RPC over HTTP with a bearer token. If your tooling speaks MCP it already speaks this, and if it does not, it is a POST with a header.

Sign in, mint a token, connect a client.

Set it up →

Your first video ad is free.

You need an account before you can mint a token, and the fastest way to get one is to make a video. Agencies get the most out of this.

Upload photos in the app