You'll noticeTier: how much it should matter to you
2Useful: my rating, 1 to 5
1Signal: worth watching, 1 to 5
SDKArea: what it touches
ImprovementsKind: in v2.1.321,
ImprovementsSection of the release
What
When Claude's reply is cut short, Claude Code reports a max_output_tokens API error. That error now has a cut field saying which limit stopped the reply.
The SDK schema adds cut with two values: output_limit and context_window. The SDK is the toolkit for building programs on top of Claude Code.
When the reply hits the output limit, the error includes apiErrorParams: { cut: "output_limit" }.
Why
Programs built on Claude Code can tell the two causes apart. A reply can stop because it hit the most text allowed in one answer (output_limit) or because the conversation ran out of room (context_window).
