Nothing to try yetTier: how much it should matter to you
2Useful: my rating, 1 to 5
3Signal: worth watching, 1 to 5
Context WindowArea: what it touches
In DevelopmentKind: in v2.1.292,
In DevelopmentSection of the release
Unclear It is not clear what this token budget limits once it is switched on.
What
Claude Code has a new setting that works out a token budget. A token is a small chunk of text, roughly a few characters, that the model reads. The budget is the size of the context window times a fraction, one sixth by default. The context window is how much text the model can hold at once. The setting also carries a figure for how many bytes count as one token.
The setting has two modes, always and after_compaction. Compaction is when Claude Code shortens a long conversation by summarising older parts of it. The setting is off unless it is switched on remotely.
Why
This is a limit tied to the size of the context window and to compaction. It does nothing until it is switched on remotely.
Read from
Feature flag
tengu_breezy_pelicanNot enough to say
Nothing here resolved what this flag was doing on this version, so nothing here should be read as on or off.
This account: no value returned · anonymous baseline: no value returned · compiled default in v2.1.292: not a boolean we can read
These values were read against a different version of Claude Code, so treat them as the nearest reading available instead of one taken on this release.
