web-read

Read a web page as clean Markdown with web-read

web-read is a CLI on $PATH that fetches a URL through Jina Reader and returns clean, readable Markdown — the page's actual content with the HTML tags, scripts, and navigation chrome stripped server-side.

Prefer web-read over raw curl when your goal is to read a page. Raw curl returns the HTML source (tags and all), which buries the content and wastes context. Use curl only for API calls or when web-read can't reach a page.

Usage

# Clean Markdown to stdout — capture it so you can grep/quote it later:
web-read "https://example.com/article" > /tmp/page.md
head -c 5000 /tmp/page.md

# Structured JSON ({title, url, content, ...}) instead of Markdown:
web-read --json "https://example.com/article"

Only http:// and https:// URLs are accepted. Quote the URL.

Behavior and exit codes

Exit	Meaning
0	Success — page content is on stdout.
2	Usage / validation error (no URL, bad scheme, unknown flag). Fix the call.
3	Fetch failed — non-2xx from Jina or a timeout. A clear error is printed to stderr. Treat as a failed fetch: try another URL or stop; do not fabricate content.

All errors go to stderr, so web-read "$url" > /tmp/page.md leaves /tmp/page.md holding clean content (or empty on failure) — never an error string mixed into the text.

Environment

Variable	Purpose
JINA_API_KEY	Optional. Sent as a Bearer token for a higher rate limit. Anonymous (no key) still works.
WEB_READ_TIMEOUT	Optional. Per-request timeout in seconds (default 20).

Privacy note

The target URL transits r.jina.ai, a third-party reader service. That is fine for public research. Do not point web-read at private, internal, or regulated content — fetch those with curl directly instead.