production/Agent log
Agent log
A plain-English account of how a video was made: which agent ran, what it decided, and why. One page per video.
What it shows
Every agent run in order, each expandable. The first is open by default. For each run you get the decision it reached and the reasoning behind it, written to be read rather than parsed.
Advanced mode
Flip Advanced on and the same log gains the machinery: model names, token counts, per-run cost and the raw reasoning stream. This is the page to open when a video cost more than you expected — it shows exactly which step spent it.
Live reasoning
While a stage is running, the pipeline screen streams the agent’s reasoning as it happens. The log is the durable record of the same thing, kept after the run finishes.
Didn’t find your answer?
Ask us directly and we’ll walk you through it — your question usually becomes the next page here.
Contact us →