Follow Discord
Sweep 22 Oct 2026 · 04:32Z Build v2.1.308 527 read Stable v2.1.308 Latest v2.1.308 Next v2.1.308 Feeds RSS JSON llms.txt llms-full.txt Unofficial

Claude Code v2.1.300 ·

PDF page reads can be capped per message and per turn once the context has filled

Two new env vars cap how many PDF pages Claude reads per message and per turn after the context has filled once

Group of 2 You'll notice Improvements
JSON All of v2.1.300
You'll noticeTier: how much it should matter to you
2Useful: my rating, 1 to 5
2Signal: worth watching, 1 to 5
Read ToolArea: what it touches
ImprovementsKind: in v2.1.300,
ImprovementsSection of the release

Unclear It is not clear what sets the enabled-and-enforced state, so setting the variables alone may not turn the limit on.

What

The context is the amount of conversation Claude can hold at once. Once it has filled up one time, reading PDF pages can now be limited:

  • CLAUDE_CODE_FILE_READ_MAX_PDF_PAGES_PER_MESSAGE sets how many pages each of Claude's messages may read. The caps only apply when this is set below the normal per-request page maximum.
  • CLAUDE_CODE_FILE_READ_MAX_PDF_PAGES_PER_TURN sets the limit for a whole turn, with a built-in default if unset.
  • A per-session record must also say the feature is enabled and enforced, with its threshold within bounds.
  • A read over the limit is refused with a message starting 'The context has filled up once already', telling Claude to read fewer pages or answer from what it has and say which pages it did not reach.

Before, the only limit was the maximum page range per single request.

Why

After a long PDF has already filled the context once, Claude is pushed to answer with what it has rather than keep re-reading pages and filling the context again.

How sure we are
Something disagreesSomething we can check disagrees with this entry, or the writer said they could not settle it.
The writer flagged doubtIt is not clear what sets the enabled-and-enforced state, so setting the variables alone may not turn the limit on.

See this entry in the whole of v2.1.300 →

Feedback