LLM token cost calculator

The biggest hidden cost in a scraping agent isn't the scrape — it's thetokens you feed the model. Raw HTML carries hundreds of bytes of tags, scripts and styles that your LLM pays for, token by token, and then has to ignore. Clean, LLM-ready markdown strips that noise.

Use the calculator below to estimate the savings on your own workload. It uses a real-world average: converting a typical HTML page to clean markdown cuts the token count by roughly2-4× — because HTML markup averages ~60-75% of a page's bytes before cleanup.

How the numbers work

A typical modern web page is ~60-75% markup by byte count (tags, attributes, scripts, styles). Clean markdown keeps the meaningful content and drops almost all of it. In practice, tokenizing the same page as raw HTML vs as markdown shows a 2-4× reduction — that's the range the calculator uses. The exact ratio varies by site; JavaScript-heavy SPAs and content sites with heavy navigation tend toward the higher end.

Why this matters for agents

If your agent reads 10,000 web pages a month, the difference between raw HTML and clean markdown isn't a rounding error. At the median ratio it's a 3× cut on the largest line item in your budget. And because Fastcrawl returns markdown by default (no extra processing step, one flat credit per call), the savings are free.