You want the top Hacker News stories in JSON. The honest options used to be: read the site’s API docs and write a client, or write a scraper and maintain it forever. AutoCLI collapses that into one command per site — Reddit, Hacker News, YouTube, Twitter/X and a long tail of others — with the output already shaped as JSON, YAML, CSV, Markdown, or a table.

It’s Rust, Apache-2.0, and the appeal is narrow but real: the ten-minute question that doesn’t justify a scraper.

Install 

  curl -fsSL https://raw.githubusercontent.com/nashsu/autocli/main/scripts/install.sh | sh

Binaries for macOS, Linux, and Windows are on the releases page, and cargo build --release works from a clone. As with any install script, read it before you pipe it into a shell — or take the binary and skip the question.

Use it 

  autocli hackernews top --limit 10
autocli hackernews top --limit 10 --format json
autocli reddit hot --limit 5 --format yaml
autocli twitter search "rust lang" --limit 10 --format csv

The shape is always autocli <site> <command> [options], and autocli <site> --help lists what a given site supports. Default output is a table for reading; --format json is what you pipe.

  autocli hackernews top --limit 30 --format json | jq -r '.[] | select(.score > 200) | .title'

That pairs it with jq the way the rest of this site expects: one tool fetches, one tool filters, neither knows about the other.

Where it earns its place

  • A morning digest. A cron job that pulls three feeds, filters by score, and mails you the result — a twenty-line shell script instead of a project.
  • Research you’d otherwise do by scrolling. Pull a week of posts to a CSV and read them in a spreadsheet.
  • Seeding a demo. Real-looking data for a prototype without writing fixtures.

Read this before you build on it

  • Site coverage is inherently fragile. Anything that reads sites rather than contracts breaks when a page changes. Fine for a script you run by hand; risky as a dependency of something that must not fail. For Hacker News and Reddit specifically, both publish official APIs — for a production job, use those.
  • Terms of service and rate limits are yours to respect. The tool will happily make more requests than a site wants. Pace it, cache what you fetch, and don’t redistribute content you don’t have the right to.
  • The browser mode is the part to think hardest about. Commands that reach logged-in content work through a Chrome extension using your own session. That’s your real account acting automatically — convenient, and worth a moment’s thought about which account, and which site’s rules on automation.

Treat it as a fast way to answer a question, not as infrastructure.

Next

On to shipping → Containers & Deploys

Edit this page

Last updated 23 Aug 2026, 00:00 UTC. history