Fetch example.com Homepage Content

Purpose

Read-only extraction of the example.com homepage and return its h1 heading text and the first paragraph text (and, optionally, the trailing "Learn more" link). example.com is the IANA-reserved illustrative domain whose homepage is a single static HTML document served by Cloudflare — no JavaScript rendering, no anti-bot, no authentication. The optimal path is a raw HTTP fetch and a minimal HTML parse; a browser session is not required.

When to Use

An agent needs a known-stable, zero-friction target to smoke-test its fetch + parse pipeline end to end.
A documentation, tutorial, or eval harness needs the canonical "hello world" web payload returned in a normalized shape.
A connectivity / DNS / TLS check needs to confirm not just that example.com is reachable but that the expected document body is being served (e.g. detecting a captive portal or middlebox interception).
A demo wants to show a JSON-shaped extraction of h1 + lead paragraph from any URL, using example.com as the safe reference input.

Workflow

The recommended path is a single HTTP fetch. example.com serves a complete, server-rendered HTML document — there is nothing for a browser to do that curl-equivalent tooling cannot.

Fetch the page with browse cloud fetch. No --proxies, no --verified, no session needed:
```
browse cloud fetch https://example.com
```
The response is a JSON envelope; the content field holds the raw HTML and statusCode should be 200.
Parse the HTML for the two required fields. The document structure is stable: a single <h1> inside a <div>, followed by two <p> elements (the descriptive paragraph and a paragraph containing only the "Learn more" <a>). Any minimal parser works — examples:
- Node: cheerio → $('h1').text() and $('p').first().text().
- Python: BeautifulSoup → soup.h1.get_text(strip=True) and soup.find('p').get_text(strip=True).
- Regex (acceptable because the document is hand-authored and stable): /<h1>([^<]+)<\/h1>/ and /<p>([^<]+)<\/p>/.
Normalize whitespace on the extracted strings (collapse runs of whitespace, strip leading/trailing) before returning. The served HTML is minified onto a single line, so naive substring extraction will not have stray newlines, but downstream consumers should still be defensive.
Return the structured shape shown in Expected Output.

Browser fallback

Only worth using if your harness has no HTTP-fetch primitive at all, or if you want a visual screenshot for a marketplace card. Cost is ~2 orders of magnitude higher than browse cloud fetch (cloud session spin-up dominates).

sid=$(browse cloud sessions create --keep-alive | jq -r .id)
export BROWSE_SESSION="$sid"
browse open https://example.com --remote
browse get markdown body --remote
# {"markdown":"# Example Domain\n\nThis domain is for use in documentation examples without needing permission. Avoid use in operations.\n\n[Learn more](https://iana.org/domains/example)"}
browse cloud sessions update "$sid" --status REQUEST_RELEASE

The browse get markdown body output is already cleanly normalized; split on \n\n to separate the heading line from the first paragraph.

Site-Specific Gotchas

The page text is not historical-museum content — it changed. The widely-quoted older version that began "This domain is for use in illustrative examples in documents…" is no longer what's served. As of Last-Modified: Thu, 14 May 2026 05:31:28 GMT, the lead paragraph reads: "This domain is for use in documentation examples without needing permission. Avoid use in operations." Do not hardcode the paragraph text in tests — extract it at runtime, or your skill will silently rot the next time IANA updates the copy.
Cloudflare edge caching is aggressive (Cf-Cache-Status: HIT, Age header in the tens of thousands of seconds is normal). The Last-Modified header is therefore the authoritative freshness signal, not Date. If you need to detect a content change, compare Last-Modified rather than re-fetching on a timer.
Allowed methods are GET, HEAD only (Allow: GET, HEAD). Do not waste retries on POST / OPTIONS; the origin will refuse them.
No robots.txt enforcement and no rate limiting observed at single-digit requests per minute. This is the IANA reference domain, intentionally permissive for documentation use. Do not abuse it (do not use it as a load-test target — there are dedicated services for that).
example.com, example.org, example.net, and example.edu all serve the same payload from the same infrastructure. If your skill is generalized for "IANA example domains", treat them interchangeably; only the host header in the request differs.
The page does NOT include the host string example.com in its visible body — only the title (<title>Example Domain</title>) and the h1 (Example Domain) name the page. Do not assume the body contains the domain literal.
There is no API in the conventional sense. The HTML document itself is the API. Do not waste iterations probing for /api/, /v1/, GraphQL, or sitemaps — none exist.
Content-Encoding: br (Brotli) is returned by default. browse cloud fetch and modern HTTP clients decode this transparently; raw socket-level clients will need to advertise Accept-Encoding: identity if they cannot decode Brotli.

Expected Output

{
  "url": "https://example.com",
  "status": 200,
  "fetched_at": "2026-05-19T00:00:43Z",
  "last_modified": "2026-05-14T05:31:28Z",
  "title": "Example Domain",
  "h1": "Example Domain",
  "first_paragraph": "This domain is for use in documentation examples without needing permission. Avoid use in operations.",
  "learn_more_url": "https://iana.org/domains/example"
}

If you cannot reach the origin (DNS failure, TLS failure, captive portal returning a non-Example Domain body), return an error shape rather than fabricated content:

{
  "url": "https://example.com",
  "status": 0,
  "error": "fetch_failed",
  "error_detail": "ENOTFOUND example.com"
}

If you reach the origin but the document shape has drifted (no h1, or zero p elements found), return a partial-success shape with the raw HTML attached for debugging — never silently substitute defaults:

{
  "url": "https://example.com",
  "status": 200,
  "h1": null,
  "first_paragraph": null,
  "error": "unexpected_document_shape",
  "raw_html": "<!doctype html>..."
}