WebFetch still can’t read claude.ai/share links (follow-up to #33985) — now returns 200 + empty shell instead of 403

Status Open
Maintainer reply None cached
Activity 0 comments · opened Oct 9, 2026

Preflight Checklist

  • [x] I have searched existing requests and this feature hasn't been requested yet
  • [x] This is a single feature request (not multiple features)

Problem Statement

WebFetch cannot retrieve content from claude.ai/share/<id> links. This was previously reported and consolidated in #33985 (itself consolidating #7162, #24462, #29009, #13843, #15881, #18803, #28307), which was auto-closed as “not planned” by the stale bot after inactivity — no maintainer ever responded, and it’s now locked, hence this follow-up.

What’s changed since #33985
Then (2026-03): curl against claude.ai/share/<id> returned HTTP 403 with cf-mitigated: challenge (Cloudflare bot challenge).
Now: The same link returns 200 OK, but WebFetch only gets the page’s static footer (Anthropic PBC’s legal/company boilerplate — registration number, Delaware incorporation, SF address, support contact, AWS hosting). None of the actual conversation content comes through.

This suggests the share page is now (or still) client-side rendered, and WebFetch — which fetches HTML and converts it to markdown without executing JS — only ever sees the pre-render shell.

For comparison

claude.ai/artifact/<id> and claude.ai/code/artifact/<id> links work correctly — but only because they’re special-cased to route through a dedicated Artifact tool (action: "read") instead of WebFetch. No equivalent special-casing exists for /share/ links.

Proposed Solution

Suggested fix

Either:

Special-case claude.ai/share/ the same way claude.ai/artifact/ already is — route to a tool that can read the real (server-side) conversation data instead of the rendered page, or
Have WebFetch (or a note in its tool description) detect this URL pattern and explicitly report “can’t be read this way” instead of silently returning irrelevant footer content, which is actively misleading — it looks like a successful fetch of an empty/unrelated page rather than a failure.
Impact

Makes it impossible to hand a Claude.ai conversation link to Claude Code for follow-up work without manually copy-pasting the full transcript — the exact workflow break described in #33985.

Alternative Solutions

Priority

Medium - Would be very helpful

Feature Category

API and model interactions

Use Case Example

Use cases for Reproduction

  1. Open any claude.ai/share/<id> link.
  2. Call WebFetch with that URL and any prompt.
  3. Result: a summary of the footer content only, never the conversation.

Reproduced consistently across multiple distinct share links in the same session.

Additional Context

_No response_

View original on GitHub ↗