Idle compaction is when Claude Code replaces older messages with a shorter summary while you are away. The two kinds of prompt cache it handles are now controlled differently:
- The one-hour cache no longer has its own remote switch. It is always on in the code.
- The keepalive cache is controlled by one remote setting with a mode. "compact" turns it on, "log" only records what would happen, and anything else turns it off. With nothing sent from the server it is off.
A minimum conversation size, read from CLAUDE_CODE_IDLE_COMPACT_MIN_TOKENS, also applies. Tokens are the small pieces of text a model counts.
Idle compaction is now rolled out differently. The one-hour cache part no longer depends on anything being switched on remotely, while the keepalive part stays off unless it is switched on remotely.
tengu_bubbly_canyon Not enough to sayNothing 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 v2.1.314: 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.
tengu_sunny_locket Not enough to sayNothing 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 v2.1.314: 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.
Read once, for one account on one subscription tier, against v2.1.314. It isn't a statement about your account. What a flag value here can and cannot tell you
A reading is one sample. Claude Code evaluates its flags remotely, so no client sees the targeting rule behind a value and this says nothing about your account.
A reading is one sample. Claude Code evaluates its flags remotely, so no client sees the targeting rule behind a value and this says nothing about your account.
A reading is one sample. Claude Code evaluates its flags remotely, so no client sees the targeting rule behind a value and this says nothing about your account.
A reading is one sample. Claude Code evaluates its flags remotely, so no client sees the targeting rule behind a value and this says nothing about your account.
A reading is one sample. Claude Code evaluates its flags remotely, so no client sees the targeting rule behind a value and this says nothing about your account.
A reading is one sample. Claude Code evaluates its flags remotely, so no client sees the targeting rule behind a value and this says nothing about your account.
A reading is one sample. Claude Code evaluates its flags remotely, so no client sees the targeting rule behind a value and this says nothing about your account.
A reading is one sample. Claude Code evaluates its flags remotely, so no client sees the targeting rule behind a value and this says nothing about your account.
New in this build: tengu_bubbly_canyon