Claude Code v2.1.237: A Concise Output Style, a Persistent Default Model, and a Silent Turn Reminder#
Part of the Claude Code Version Tracker series. | Official Env Vars | Official Changelog
Coming from v2.1.234, a v2.1.237 user can pick a built-in "Concise" output style, pin the model new sessions start on without losing the ability to switch, and, on macOS and Linux, ask one Claude Code session to ping another when it goes idle. Three published builds (v2.1.235, v2.1.236, v2.1.237) land between the two versions, and the binary confirms a fourth change the notes skip.
The Concise style is selectable under Output style in /config.[1] Its system text is in the binary: "Answer simple questions in 1-3 sentences of plain prose," "Don't restate the request, the plan, or each step you took," with a guard that "Conciseness never means withholding requested information" and that "Error reports, failing test output, security warnings, and confirmations for destructive actions keep their full content." So it trims narration and preamble while leaving diagnostic and safety output intact.
ANTHROPIC_DEFAULT_MODEL (shipped v2.1.236, 2026-08-19) sets the starting model for new sessions.[2] A /model pick still overrides it and persists across restarts, which is the difference from ANTHROPIC_MODEL. In the binary the model picker labels a value from this variable as "Set by ANTHROPIC_DEFAULT_MODEL," distinct from the org-enforced "Org default" row.
The turn reminder the notes skip#
CLAUDE_CODE_SILENT_TURN_REMINDER and CLAUDE_CODE_SILENT_TURN_REMINDER_TURNS first appear in v2.1.236, and that release's changelog does not mention them. When enabled, after a run of turns with no word to the user Claude receives an injected line: "The user hasn't heard from you in a while. As you continue, keep them updated when there's something to tell — a finding, a change of plan." The turn count falls back to the tengu_hushed_lark gate and defaults to 5. It is a nudge for long autonomous stretches to surface progress instead of going quiet.
| Variable | What It Does |
|---|---|
ANTHROPIC_DEFAULT_MODEL | Sets the model new sessions begin on. A /model selection overrides it and is remembered across restarts, unlike ANTHROPIC_MODEL. The picker marks a value set this way separately from an organization-enforced default. Shipped v2.1.236. |
CLAUDE_CODE_SILENT_TURN_REMINDER | Toggles an injected reminder that fires after several turns with no update to the user, prompting Claude to report a finding or a change of plan. Overrides the server-side default. Present since v2.1.236, unmentioned in that release's notes. |
CLAUDE_CODE_SILENT_TURN_REMINDER_TURNS | Sets how many silent turns pass before the reminder fires. Falls back to the tengu_hushed_lark gate, defaulting to 5 turns. |
Cross-session, input, and rendering fixes#
SendMessage gains notify_when_idle: ask another Claude Code session on this machine to send one notice when it next goes idle, opt-in and one-shot with no polling (macOS and Linux).[2] The bundled tool text warns that the notice is not an answer to any pending question. An optional spellcheck setting underlines misspelled words in the prompt as you type, using an installed aspell, hunspell, or ispell.[3] Prompt caching now works for sessions routed through an LLM gateway or a custom base URL.[1]
Two robustness fixes are worth noting. On macOS, wildcard read-deny sandbox rules such as **/.env now take precedence inside allowed read regions, cover a matched directory's contents, and survive a rename of the denied file.[2] The fullscreen renderer, after a failed start, falls back to the classic renderer instead of exiting on every later launch.[2]
What these tell us#
The two headline knobs pull in the same direction: less friction per session. A default model that sticks until you deliberately change it, and an output style that drops narration while keeping error and security text, both reduce what a user has to manage or re-read each turn.
The silent turn reminder is the release's most complete unannounced piece. It targets long autonomous runs, the same stretches where the default model and concise output matter most, by having Claude check in when it has been heads-down for a set number of turns. It ships gated with an environment-variable override and a tunable threshold, in place ahead of any changelog line, consistent with earlier releases staging autonomy controls before they are announced.
This analysis is conducted for educational and research purposes under fair use principles. All trademarks and software referenced belong to their respective owners. This content is not intended to infringe on any intellectual property rights, circumvent any protections, or encourage unauthorized access to proprietary systems.
Sources#
- Claude Code Official Changelog, v2.1.237 release notes
- Claude Code Official Changelog, v2.1.236 release notes
- Claude Code Official Changelog, v2.1.235 release notes
Related Versions#
- Claude Code v2.1.221: Sandbox Credential Masking, a Bash Permission-Bypass Fix, and Silent Artifact Comment Threads. Credential masking, MCP protocol negotiation, artifact comment threads.
- Claude Code v2.1.219: Opus 5 Becomes the Default Opus Model. New flagship, native 1M context, per-machine dispatch switch.
- Claude Code v2.1.218: Quieter Reviews, a Guard on Team Memory, and Two Knobs the Notes Skip. Review noise controls, memory write guard.
- CLAUDE_CODE_EXPERIMENTAL_OBSERVER_AGENTS: The Undocumented Observer Agents Flag. Background observer paired to an agent, shipped silently in v2.1.200.