read
scrollhound read <url> [--comments <n>] [--cursor <cursor>]The one command to reach for when someone shares a link. It works out what the link is and returns all of it:
| Link | What you get |
|---|---|
| X post or thread | The author’s whole thread in order (long posts in full, media, quoted posts) |
| X Article | The Article as Markdown (long ones are split into chunks) |
| YouTube video | Title, channel, views and the full transcript |
| Hacker News item | The story or comment |
Reddit post (reddit.com/r/…/comments/…, redd.it/…) |
The post, as reddit post |
Subreddit (reddit.com/r/…) |
Its description, size and rules, as reddit sub |
scrollhound read https://x.com/levelsio/status/1849000000000000000scrollhound read https://www.youtube.com/watch?v=dQw4w9WgXcQscrollhound read https://news.ycombinator.com/item?id=8863scrollhound read https://www.reddit.com/r/SaaS/comments/1xaofm2/Comments and replies aren’t included unless you ask for them with --comments <n> (X keeps up to
100, Hacker News 200, Reddit 500). The agent asks when you want the discussion or reactions, or
when the task needs them. Without it, the read ends with a line such as
318 comments on Reddit, not loaded (ask for comments to include them).
scrollhound read https://news.ycombinator.com/item?id=8863 --comments 30Reddit share links (reddit.com/r/…/s/…) can’t be read directly: open one in a browser and pass the
full post URL instead.
read doesn’t transcribe X videos by itself; it marks them with
[video 1:42] transcript available: scrollhound transcript … so the agent can ask for one when it
matters. See transcript.
When there’s more (another page of the next chunk of a long transcript), the output ends
with a next: line holding the exact command to run.