Claude
Sonnet 5 — $2.00 / MTok input
$24.80
saved per 1,000 reports — $51.20 as HTML, $26.40 as Markdown.
Save tokens and improve accuracy by converting rich-text files into an agent-driven format.
Drop a file here
The first 10 converts are free, no credit card required.
Why Markdown
A model doesn't read your document — it reads tokens, roughly four characters each. Every tag, style attribute, table cell wrapper, and layout artifact inside a PDF or Word file becomes tokens too, and you pay for all of them before a single sentence of meaning arrives.
Markdown carries the same structure in plain characters. A heading is two symbols instead of a nested element; a table is pipes instead of thirty tags. The document shrinks, so the same context window holds more of it, and each request costs less.
Accuracy improves for the same reason. When headings, lists, and tables survive as unambiguous text, an agent stops guessing where a section starts and what a column means.
HTML source
<h2 class="sec-title" id="revenue">Q3 revenue by region</h2> <table class="tbl data-table"> <thead> <tr><th scope="col">Region</th><th scope="col">ARR</th><th scope="col">YoY</th></tr> </thead> <tbody> <tr><td>North America</td><td>$12.4M</td><td>+18%</td></tr> <tr><td>EMEA</td><td>$4.1M</td><td>+31%</td></tr> <tr><td>APAC</td><td>$2.8M</td><td>+44%</td></tr> <tr><td>LATAM</td><td>$0.9M</td><td>+12%</td></tr> <tr><td>Japan</td><td>$1.6M</td><td>+7%</td></tr> </tbody> </table>
512 characters≈ 128 tokens
Markdown
## Q3 revenue by region
| Region | ARR | YoY |
| ------------- | ------ | ---- |
| North America | $12.4M | +18% |
| EMEA | $4.1M | +31% |
| APAC | $2.8M | +44% |
| LATAM | $0.9M | +12% |
| Japan | $1.6M | +7% |262 characters≈ 66 tokens
Same five regions, same ARR and growth figures — 48% fewer tokens for an agent to read.
What that saves
Take the table above as one section of a 40-page report — about 25,600 input tokens as HTML, 13,200 as Markdown. Run a thousand of those reports through an agent and the difference shows up on the invoice.
Claude
Sonnet 5 — $2.00 / MTok input
$24.80
saved per 1,000 reports — $51.20 as HTML, $26.40 as Markdown.
OpenAI
gpt-6-astra — $10.00 / MTok input
$124.00
saved per 1,000 reports — $256.00 as HTML, $132.00 as Markdown.
Gemini
3.8 Flash — $0.75 / MTok input
$9.30
saved per 1,000 reports — $19.20 as HTML, $9.90 as Markdown.
Token counts estimated at four characters per token. Rates from Anthropic's, OpenAI's, and Google's published API pricing, September 2026; all are subject to change. Gemini 3.8 Flash input is $0.75 / MTok through 31 December 2026. Claude and OpenAI are trademarks of their respective owners.
Supported formats
Text, headings, and tables are recovered from the page layout, including multi-column reports.
Documents, decks, and spreadsheets convert to headings, lists, and pipe tables.
Saved pages and exported HTML lose their wrappers and keep the content.
Files up to 50 MB. Conversion happens in your session; nothing is stored afterwards.
Convert it, read it in Annotate, and keep the Markdown. Nothing to install.
Convert for free. We won't keep the file on the server.