Bring your own keys: what the model bill really looks like

Most AI marketing tools sell you credits. A credit is a unit invented so that you cannot compare their price to the provider's price. ZwayRank does the opposite: you paste your own OpenRouter key for the models and your own DataForSEO key for search data, the app calls them on your behalf at cost, and it shows you every cent by brand and by agent. This piece is what those cents look like.

The three lines on the bill

Tracking. Every day the core questions are asked on each engine, and once a week the full set. One question on one engine is a fraction of a cent on a fast model and a few cents on a strong one. For a brand with forty core questions on three engines, the daily run is usually under a dollar, and the weekly full scan a few dollars. This is the biggest line for most workspaces, roughly forty percent of the total.

Writing and images. A brief costs a few cents. An article on a strong model costs somewhere between twenty and sixty cents. Two images add a similar amount depending on the image model. Call it under a dollar a page.

Social visuals. A still is a few cents. A carousel is a handful of stills. A short clip is the expensive one, from tens of cents to a few dollars depending on the video model. That is why social has four tiers, from text only to the full mix, and why the tier is yours to pick per brand.

Search data is the fourth line and it is small: keyword volumes and rankings for a brand, weekly, is about a dollar or two on DataForSEO.

What a month looks like

From real workspaces in 2026, with the full crew on:

• One brand, light mode, tracking only: under $10 a month.
• One brand, full crew, text-only social: $10 to $25.
• Two brands, full crew, stills for social: $20 to $70.

Your numbers depend on how many questions you track, which models you route to, and how much the crew is allowed to write. All three are switches.

One cap, blended

The problem with "at cost" is that costs are spread across providers and models. ZwayRank blends them into one monthly cap for the workspace. Before every paid call it checks the cap. A call that would break it is refused, and the record says so; nothing is silently cut short or degraded. When the cap is hit, the crew keeps doing the free work (reading, planning, healing, publishing what is already approved) and stops the paid work until the month turns or you raise the cap.

Why we do it this way

Because a metered tool has an incentive to meter. We would rather be paid once for the software and have the model bill be your business with your provider. It also means the number in the app is the number on your invoice, reconciled against the provider, so a founder can trust it without trusting us.

The first scan is on us

A new brand's first scan runs on our keys, in light mode, so you see a real report before you paste anything. After that, two keys, one cap, and the budget screen does the rest.