jurl

curl that reads the page for you.
Tell it what you want. It picks it out of the page.

$ jurl -q "how do I install it on macOS?" github.com/BurntSushi/ripgrep
…
### Installation

…

If you're a macOS Homebrew or a Linuxbrew user, then you can install ripgrep from homebrew-core:

```
$ brew install ripgrep
```

If you're a MacPorts user, then you can install ripgrep from the official ports:

```
$ sudo port install ripgrep
```

$ jurl --find "a cathedral" en.wikipedia.org/wiki/Cologne
https://thumb.wikimedia.org/wikipedia/commons/thumb/2/27/Kdom.jpg/250px-Kdom.jpg

$ jurl --links -n 3 news.ycombinator.com
https://github.com/PowderworksCode/headstart
https://www.da.vidbuchanan.co.uk/blog/hacking-time.html
https://gamehistory.org/5k-magazines/

Why jurl

• It picks. It doesn't write. jurl doesn't use a chatbot. It uses decision models: Jev reads the text and Clef looks at the images. Neither writes a word. They only say how likely each paragraph, link or image is to be what you want, and jurl prints the winners exactly as they appear on the page. You never get a summary that drifts or a URL that doesn't exist.
• It's fast. Most pages take under a second, end to end.
• It's cheap. About a thousand pages per dollar of API usage.

Install

brew install rayoplateado/tap/jurl                                                   # macOS, Linux
curl -LsSf https://github.com/rayoplateado/jurl/releases/latest/download/jurl-installer.sh | sh   # no Homebrew

_{Windows: powershell -ExecutionPolicy Bypass -c "irm https://github.com/rayoplateado/jurl/releases/latest/download/jurl-installer.ps1 | iex" · From source: cargo install --git https://github.com/rayoplateado/jurl}

Then run it. The first time, jurl asks for a TypeSafe API key, checks it and saves it. That's the whole setup. Pages that need JavaScript just work too: jurl fetches a headless browser the first time one shows up.

To use --vision and --find, run jurl init and add a Cloudflare Workers AI token.

What you can do

Get the gist

$ jurl -n 3 blog.cloudflare.com/markdown-for-agents/
# Introducing Markdown for Agents

<https://blog.cloudflare.com/markdown-for-agents/> · article (1.00)

The way content and businesses are discovered online is changing rapidly. In the past, traffic originated from traditional search engines, and SEO determined who got found first. …

## Convert HTML to markdown, automatically

Cloudflare's network now supports real-time content conversion at the source, …

Here’s how it works. To fetch the markdown version of any page from a zone with Markdown for Agents enabled, the client needs to add the **Accept** negotiation header …

You get the title, what kind of page it is, and the blocks that carry it, in reading order. Navigation, cookie banners, sign-up prompts, author bios and footers are left out.

Ask one question

$ jurl -q "What is the Jevons paradox?" -n 2 en.wikipedia.org/wiki/William_Stanley_Jevons
…
Jevons received public recognition for his work on The Coal Question (1865), in which he called attention to the gradual exhaustion of Britain's coal supplies and also put forth the view that increases in energy production efficiency leads to more, not less, consumption. …

## Practical economics

In The Coal Question, Jevons covered a breadth of concepts on energy depletion …

Out of a page with about 240 blocks, you get the two paragraphs that answer it.

Just the code

$ jurl --code -n 2 github.com/BurntSushi/ripgrep
### Installation

```
$ brew install ripgrep
```

```
$ cargo install ripgrep
```

Links worth following

$ jurl --links -n 3 -q "official installation instructions" github.com/BurntSushi/ripgrep
https://www.macports.org/ports.php?by=name&substr=ripgrep
https://packages.gentoo.org/packages/sys-apps/ripgrep
https://chocolatey.org/packages/ripgrep

The ranking puts content first, ahead of login, share or privacy policy. Add -q to keep only the links about something specific.

The real images

--image picks the content images by their file name, alt text and caption, with no logos, icons or tracking pixels. --vision also has Clef look at the pixels, which matters when the alt text says nothing:

Find a photo by description

$ jurl --find "a cathedral" -n 3 en.wikipedia.org/wiki/Cologne

The page has 73 images, and Clef looks at every one of them in parallel. It took 2.3 seconds:

If nothing matches, jurl tells you so instead of handing you the least-bad photo:

$ jurl --find "a carnival parade" en.wikipedia.org/wiki/Cologne
jurl: no image in https://en.wikipedia.org/wiki/Cologne looks like "a carnival parade" (closest: …, p=0.02)

JavaScript apps

Some pages arrive empty because their content is built by JavaScript. jurl spots those and renders them in Lightpanda, a fast headless browser:

$ jurl -n 2 hn.algolia.com
jurl: no text without JavaScript, rendering with Lightpanda…
# HN Search powered by Algolia

<https://hn.algolia.com/> · listing (1.00)

Stephen Hawking has died(http://www.bbc.com/news/uk-43396008)

6015 points|Cogito|9 years ago|436 comments

Use -r to force it.

Recipes

Everything prints plain text or plain URLs, so jurl fits in a pipe:

# Read the top 3 Hacker News stories, one key paragraph each
jurl -l -n 3 news.ycombinator.com | xargs -n1 jurl -n 1

# Download the photo that matches
jurl -f "a bridge over a river" en.wikipedia.org/wiki/Cologne | xargs curl -sO

# Every chart in a post
jurl -f "a chart or graph" -n 10 blog.cloudflare.com/markdown-for-agents/ | xargs -n1 curl -sO

# Only the confident answers, for scripts
jurl --json -q "installation" github.com/BurntSushi/ripgrep | jq -r '.blocks[] | select(.p > 0.8) | .text'

Reference

Keys live in ~/.config/jurl/env. Environment variables take precedence over that file: TYPESAFE_API_KEY, and for images CLOUDFLARE_ACCOUNT_ID plus CLOUDFLARE_AI_TOKEN.

Speed and cost

Measured on 2026-10-04. Run any command with -t to see your own numbers.

Your keys, your data

jurl has no server and no account of its own. It talks to the model APIs directly with your keys:

• Who you pay: usage is billed by TypeSafe (Jev, $0.042 per million input tokens) and Cloudflare (Clef-flash, $0.09 per million). Output is free on both.
• What leaves your machine: the text of the page goes to TypeSafe. With --vision or --find, the images go to Cloudflare too. Keep that in mind for internal or private pages.
• What jurl can't read: it sends no cookies, so pages behind a login are out of reach.

Update and uninstall

• Update: brew upgrade jurl, or run the install script again.
• Uninstall: brew uninstall jurl, or delete ~/.local/bin/jurl. To remove everything, also delete ~/.config/jurl (keys) and ~/Library/Caches/jurl or ~/.cache/jurl (the browser).

url ─▶ fetch (asks for markdown first) ─▶ split into blocks · links · images
          └─ empty JavaScript app? ─▶ render in Lightpanda
                          │
            one yes/no question per candidate
              ┌───────────┴───────────┐
              ▼                       ▼
       Jev reads text          Clef looks at images
   "is block 12 the point?"   "does this show a cathedral?"
          p = 0.97                  p = 0.96
              └───────────┬───────────┘
                          ▼
        rank · threshold · page order · print verbatim

cargo test        # extraction: layout tables, lazy images, markdown, links, app-shell detection
cargo build --release && ./target/release/jurl -t <url>

Contributing

Issues and pull requests are welcome. Read CONTRIBUTING.md first; security problems go to SECURITY.md.

License

MIT or Apache-2.0, at your option. Lightpanda, which jurl downloads for JavaScript pages, is a separate program under AGPL-3.0.