@m4ss/pi-web-fetch

Web-fetch tool for Pi: fetches HTTP(S) URLs, converts HTML to clean Markdown, and pretty-prints JSON responses

Packages

Package details

extension

Install @m4ss/pi-web-fetch from npm and Pi will load the resources declared by the package manifest.

npm repohome report
$ pi install npm:@m4ss/pi-web-fetch
Package
@m4ss/pi-web-fetch
Version
0.2.0
Published
Oct 28, 2026
Downloads
224/mo · 212/wk
Author
paolobarbolini
License
MIT
Types
extension
Size
34 KB
Dependencies
2 dependencies · 4 peers
Pi manifest JSON
{
  "extensions": [
    "./src/index.ts"
  ]
}

Security note

Pi packages can execute code and influence agent behavior. Review the source before installing third-party packages.

README

@m4ss/pi-web-fetch

We built this so the model can fetch web pages and get clean Markdown instead of raw HTML. It converts HTML automatically, blocks private network addresses, and caps response size to keep things manageable.

Features

  • Fetches any HTTP/HTTPS URL and converts HTML to clean Markdown (JSON is pretty-printed)
  • Asks for text/markdown first, so sites that publish Markdown for agents skip the HTML conversion
  • Blocks private network addresses by hostname (localhost, RFC 1918, link-local, CGNAT, IPv6 ULA); the initial URL and every redirect hop are re-validated — see Private networks for scope and the DNS-rebinding limitation
  • Follows instant meta-refresh redirects (the stub pages static site generators emit)
  • 10s default timeout, overrideable per call; large responses capped at 5MB

Install

pi install npm:@m4ss/pi-web-fetch

Usage

web_fetch(url="https://www.postgresql.org/docs/current/", format="markdown")
web_fetch(url="https://redis.io/docs/latest/", format="html")
web_fetch(url="https://api.github.com/repos/M4SS-Code/watermelon/contents")
web_fetch(url="https://example.com/slow", timeoutMs=20000)

Private networks

The initial URL and every redirect hop (HTTP 3xx or meta-refresh; max 5, each re-validated) is checked by hostname and refused when it names a private or internal address: localhost and other single-label hostnames, RFC 1918, loopback, link-local (including the 169.254/16 cloud-metadata range), CGNAT, and IPv6 loopback/link-local/ULA. Numeric host forms are normalized first, so http://0x7f000001/ and http://2130706433/ are refused too.

The check runs on the URL's hostname, not on the resolved address, so it does not defend against DNS rebinding: a hostile public hostname that resolves to a private address gets through (closing that would require a pinned resolver that validates and uses the resolved IP). Treat the guard as protection against accidental internal access, not as a hard security boundary; don't rely on it to isolate a machine that can reach sensitive internal services.

Related packages

License

MIT