Claude Code now records the permission mode a tool request was made under (permissionModeAtRequest) to resolve 'auto' mode and keep decisions consistent
What
- A new
permissionModeAtRequestfield is now threaded through tool-use handling, capturing the concrete permission mode (default,acceptEdits,dontAsk, orbubble) that an "auto" mode resolves to at the moment a tool request is made;planandbypassPermissionsboth resolve todefault. This replaces an earlier, simpler mode field. - When replying to a tool_use request, Claude Code now restores the permission mode that request was originally made under, but only when the effective mode is "arbiter" and a gate check passes.
- A new telemetry event,
tengu_auto_mode_decided_as_requested_mode, fires when auto-mode's eventual decision matches the mode that was originally requested, recording the tool name, whether it's an MCP tool, and the requested mode.
Why
This keeps permission decisions consistent with the mode that was actually in effect when a tool request was made, and provides telemetry to measure how often auto-mode's decision simply confirms the original request.
tengu_radiant_teapot 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 v2.1.275: on
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 v2.1.275. It isn't a statement about your account. What a flag value here can and cannot tell you
The gate `tengu_radiant_teapot` controlling this has no recorded reading, so whether it is active for any account is unknown.
New in this build: tengu_radiant_teapot