A hostCredential permission source and a reserved consent envelope key are in place, with nothing yet setting or reading them.
Approval-related plumbing gained a distinct 'withdrawn' outcome: a new user-facing message is shown when an approval answer arrives for a request that was already withdrawn, and a new set of terminal reasons (no_answer, host_withdrawn, superseded) plus a consent field were added to the tracked fields for tool-use records.
The permission-decision source enum gained a hostCredential entry (joining session, toolsNarrowing, mcpServerPolicy), and the reserved tool-envelope keys list gained consent alongside tool/tool_use_id, with envelope-building code now explicitly stripping a consent field into the shadowed-keys bucket. This looks like scaffolding for a host-mediated tool-consent/credential flow; no in-bundle caller was found that actually sets or reads a consent value on a tool envelope, nor a UI path that surfaces the withdrawal message.
The entry above is what we published on the day. These lines were added later, as Anthropic's own pages caught up, and they sit beside the original rather than replacing it.
* Fixed long-context 429s on Fable models showing the usage-credits consent prompt instead of the 1M-context message on Pro and Team planschangelog see the edit
Anthropic's documentation has since written up consent, on Claude Code changelog.