Site to Markdown
Turn a docs site into clean Markdown with source links, ready for an AI to read.
under 5 runs so farWorked: not enough data yetneeds 20 runs from 3 people
Built by Modpack on trafilaturaOpen-source project (Apache-2.0). Fixes we make go back upstream.
People ask
$0.01per first 10 pages
$5 to start, about 496 first 10 pagesAdd to ClaudeLink didn’t open?
Answer · from the pack’s docs and checks
Does it keep code blocks and tables?
Yes. Code keeps its language and tables become Markdown tables.
Pack docsPeople ask
Sample result
Your example (samples/quick.input.json) · first 10 pages{
"schema": "modpack.result/v1",
"pack": "modpack/site-to-markdown",
"entrypoint": "quick",
"status": "ok",
"subject": "demo.modpack.egoeimi.org",
"headline": "demo.modpack.egoeimi.org: 9 pages converted to Markdown, about 400 tokens",
"summary": "Converted 9 pages of demo.modpack.egoeimi.org into Markdown with a source link on each: 301 words, about 400 tokens, small enough for one conversation. 6 code blocks kept their language and 1 table stayed a table. Skipped: 1 page couldn't be read. The quick run reads up to 10 pages; the site has about 10.",
"stats": [
{
"label": "pages converted",
"value": "9"
},
{
"label": "words",
"value": "301"
},
{
"label": "tokens (estimate)",
"value": "400"
},
{
"label": "pages on the site (estimate)",
"value": "10"
}
],
"table": {
"title": "Pages, in reading order",
"columns": [
{
"key": "page",
"label": "Page"
},
{
"key": "title",
"label": "Title"
},
{
"key": "words",
"label": "Words"
}
],
"rows": [
{
"page": "/",
"title": "Know when your jobs stop",
"words": 20
},
{
"page": "/docs/intro",
"title": "Introduction",
"words": 80
},
{
"page": "/docs/install",
"title": "Install",
"words": 29
},
{
"page": "/docs/configure",
What changed
Facts and evidence
- Rebuilt from scratchSep 28
- Checks passedSep 28
- Works with
- Claude Code
- You give it
- The docs site or page to start from, e.g. https://docs.example.com. A bare domain works too.; Pages to convert (the quick run reads at most 10).; Stay inside this path, e.g. /docs/. Default: the start page's first folder when it isn't the home page.
- You get
- result; site; index; pages; pages_jsonl; crawl
- Needs
- Nothing to connect
- Time
- About 20 s for a first 10 pages
- Cost
- $0.01 per first 10 pages · $0.01 per whole site. You pay from your Modpack balance, only when a run works.
- Worked
- Not enough data yet (a rate shows after 20 runs from 3 people in 30 days)
- Median run
- 20 s in the last 30 days
- Limits
- Public pages only. Pages behind a login, paywall or consent wall that hides content aren't read. No browser runs, so JavaScript isn't executed. Pages that build their text in the browser come out empty or short; they're skipped and listed. robots.txt is honored for the ModpackBot crawler; disallowed pages are skipped and listed. The crawler makes at most 4 requests a second and stops at the page or time limit. Only the site's own pages are converted. Linked PDFs and other files are listed, not converted. The page count on the site is an estimate from the sitemap and the links the crawl saw. E-mail addresses in page text are replaced with "(email address removed)"; bylines and author names are dropped with the rest of the page chrome. Code blocks are kept exactly. site.md stops at 5 MB; every page is still in `pages/` and `pages.jsonl`. Tokens are estimated as words × 1.33; real counts depend on the model.
- Version
- v1.4.0 · Sep 28 · new versions pass the same checks first
- Last verified
- Sep 28
Related packs
How it works
- 01Add to ClaudeOnce. Then it’s in Claude on web, desktop and phone.
- 02Ask in your words“Use Modpack’s Site to Markdown pack”
- 03Get the resultRight in the chat, with the details and files.