News

Runway Agent Now Opens Files and Runs Code, and Runway Ads Closes the Loop to the Ad Platforms: What Runway Says, and What It Has Not Said

1 November 2026 · RenderBob team

Runway says its Agent now does more than generate: it can open files, run code and build a brand launch kit. Separately, Runway Ads generates ad creative, publishes it and reads results back.

Runway says its Agent now does more than generate: it can open files, run code and build a brand launch kit. Separately, Runway Ads generates ad creative, publishes it and reads results back. Both are described mainly by Runway.

What Runway Agent is said to do

A Creators Toolbox post on 30 October, relaying Runway's announcement on X and a 64-second demo video, lists:

• Opening files and running code.
• Building a brand launch kit from a website address.
• Making product shots, ad variations and motion graphics from a reference.
• Writing code for exact placement, such as a logo at a set margin.

The list ends with "and more", so it is not complete. The post says it could not verify the details against Runway's product documentation. The Agent page says Agent is offered across Runway's plans, but the post does not tie that to these new capabilities, and the changelog's newest entry on 30 October was an earlier note about Runway Ads. No pricing or rollout timing has been given.

What Runway Ads is said to do

Runway's own page, dated 30 September, calls it "an autonomous engine for performance marketing". It says it:

• Makes video and image ads from brand guidelines, past ads and product imagery.
• Publishes approved variants to Meta, Google and TikTok, reads performance data back and makes the next round.
• Localises assets, including on-screen text and product screenshots, under team-set rules.
• Runs an automated brand check before approval, with human approval on by default.
• Enforces budget limits such as a maximum daily change.

Runway reports its own results since July: weekly ad volume from 77 to roughly 900, return on ad spend doubled, conversion up about 34% with click-through steady, and cost per subscriber down 41%. These are Runway's internal figures. The page gives no pricing or availability.

What these claims are worth

The Agent description is a demo and a social post. The Ads results come from the vendor, with no named customer, baseline or time window, and no independent test. The sensible reading is that Runway is moving from a tool that makes clips to one that carries a job from brief to published asset. Whether the output holds up is untested.

Questions to ask before handing over a campaign

• Where does a human approve? Ads says approval is on by default. Ask what changes when you turn it off.
• What does the agent read and keep? Files, brand sources and performance data all pass through it.
• What happens to likeness and rights? See our 28 October piece on regional remakes.
• How are results measured? Ask for the baseline behind any "doubled" figure.
• Can you see each version? You need a record of what ran, where and when.
• What does it cost per usable ad, not per generation?

What to watch

Runway's own documentation for the new Agent abilities, pricing and plan limits, independent results from Ads, and whether other platforms follow with publish-and-measure loops.

RenderBob 3 and this story

RenderBob 3 keeps the human in the loop: generated clips sit on their own lane, the model is chosen per clip, and references and prompt logs stay with the studio, so an agent's first pass is something you review and keep a record of.