@pi-lab/webfetch

WebFetch tool extension for pi coding agent

Package details

extension

Install @pi-lab/webfetch from npm and Pi will load the resources declared by the package manifest.

$ pi install npm:@pi-lab/webfetch

Pi manifest JSON

{
  "extensions": [
    "./dist/index.mjs"
  ]
}

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

A web fetching extension for pi coding agent. Adds a webfetch tool — fetch any URL and get back clean Markdown, ready for the model to read.

pi install npm:@pi-lab/webfetch

HTML → Markdown via Mozilla Readability (same engine as Firefox Reader Mode) + Turndown. Falls back to full-page conversion if Readability can't extract a main article.
Pagination — large pages are sliced into chunks; the model reads page by page using offset.
Inline script index — <script> tags are stripped from the Markdown body but listed as a numbered index at the end. The model can read any of them with script=N.
Redirect handling — same-domain redirects are followed automatically (up to 10 hops); cross-domain redirects are surfaced to the model so it can decide whether to follow.
Binary downloads — non-text responses (PDFs, images, etc.) are saved to ~/.pi/agent/pi-lab/tmp/webfetch/ and the file path is returned.
LRU cache — processed Markdown is cached in memory so paginating the same URL doesn't re-fetch.