Back to rankings

Claude Code WebFetch Now Reads Past 100,000 Characters

Claude CodeView original changelog

Claude Code 2.1.290 fixes WebFetch silently dropping page text beyond 100,000 characters. The tool now reports how much was left unread and accepts an offset parameter so Claude can read on.

Key Takeaways

  • Silent truncation at 100,000 characters is gone: WebFetch now states how much text was left unread.
  • A new offset parameter lets Claude continue reading a long page from where the previous call ended.
  • Long documentation pages can now be read in full across several calls.
  • Answers built on web pages are less likely to rest on a partial read nobody knew about.
  • No settings change is needed; the behavior applies to the existing WebFetch tool.
  • The fix is in 2.1.290, released 2026-10-05.

The problem

WebFetch truncated page text past 100,000 characters without telling Claude or the user. Long documentation pages could therefore be summarized from a partial read, with no sign that anything was missing.

The fix

In Claude Code 2.1.290, WebFetch says how much of the page was unread and takes an offset parameter to continue from where it stopped. Claude can page through a long document in several calls instead of working from a silently cut-off view.