Unclear It is unclear when the REPL clean-up runs, so whether subagents actually use their own REPL sessions in this release is not settled.
A forked agent is a separate copy of Claude that Claude Code starts to handle part of a task. Claude Code can now start one with a pinned effort, a fixed level of how hard the model thinks, which is passed through to the agent's requests.
When a forked agent finishes, Claude Code can now also close the agent's REPL session (an interactive environment where code is run step by step). This happens only when an internal check allows it. Before, clean-up only stopped the agent's shell tasks. If closing the REPL session fails, Claude Code logs REPL session end failed (continuing teardown) and carries on cleaning up.
The separate relay-fork switch is still off unless it is switched on remotely. At its most recent reading, taken before this release, the flag server returned it on for this site's account and for the anonymous baseline.
This suggests subagents (helper copies of Claude working on part of a task) may run REPL sessions of their own, which now get closed properly when the agent ends.
tengu_wild_tome Off by default, switched on for this accountThe shipped code defaults this off, and the flag server returned on for the one account this site reads on this version. That is the reading that makes the entry above worth a second look, and it still says nothing about your account.
This account: on · anonymous baseline: on · compiled default in v2.1.318: off
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 v2.1.318. It isn't a statement about your account. What a flag value here can and cannot tell you
It is unclear when the REPL clean-up runs, so whether subagents actually use their own REPL sessions in this release is not settled.