Feature flags, all 2,094.

tengu_cedar_marsh ◇ Default false. When on, reduced motion is applied in an xterm.js-based terminal such as VS Code's integrated terminal (detected from VS Code terminal markers, TERM_PROGRAM=vscode, or an XTVERSION reply starting with xterm.js) even when the prefersReducedMotion setting is off. When the setting is on, reduced motion applies regardless of the flag.v2.1.157 → v2.1.309 ◆in the build never read tengu_cedar_sundial ◇ Default false. When on, an edit to a file that changed on disk since it was last read is allowed to proceed if the edit's old_string still matches the current contents exactly once (or matches at all with replace_all), instead of failing with File has been modified since read. A missing or ambiguous match still fails.v2.1.178 → v2.1.207 ○gone AND v2.1.308 tengu_cheerful_horizon ◇ Default false. When on, Claude Code stops requesting whether the claude.ai account has GitHub connected. It uses only a status it already holds, such as the one cached in config as githubWebConnectionStatusCache for the same account and organization, and a live check answers unknown.v2.1.259 → v2.1.309 ◆in the build AND v2.1.308 tengu_chrome_tool_call_disconnected ◇ The browser integration now reports connection and mismatch events, including tool calls attempted while disconnected.never mined ·never mined never read tengu_chrome_tools_added ◇ Running the Chrome extension's MCP server now logs how many browser tools were added, once per session.never mined ·never mined never read tengu_cinder_heron ◇ Questions left open when the stream closes could stay answerable, but that behaviour is switched off for you today.v2.1.222 → v2.1.235 ○gone never read tengu_cinder_relay ◇ A built-in commit command with a strict git workflow exists but is off unless a server enables it.never mined ·never mined never read tengu_classifier_summary_kill ◇ No description from anybody.v2.1.138 or earlier → v2.1.309 ◆in the build never read tengu_claude_api_skill_loaded ◇ The Claude API skill reloads its documents each time instead of caching the load.never mined ·never mined never read tengu_claudeai_mcp_eligibility ◇ An analytics event, not a feature flag. Claude Code logs it while deciding whether to load claude.ai MCP connectors, with a state field: disabled_env_var, disabled_setting, safe_mode, third_party_provider, api_key_precedence, no_oauth_token, missing_scope, eligible after the server list is fetched, or fetch_failed with the HTTP status and the number of attempts.never mined ·never mined never read tengu_claudeai_skills_sync_idle_intervals ◇ No description from anybody.v2.1.293 → v2.1.308 ○gone never read tengu_claudeai_skills_sync_interval_minutes ◇ Code default 10. Sets, in minutes, how long the background sync of claude.ai skills waits before running another round. A value that is not a number, or is zero or less, falls back to 10 minutes, and other values are clamped to between 5 and 1440. After consecutive failed rounds the wait doubles each time, up to four times the interval.v2.1.282 → v2.1.309 ◆in the build never read tengu_claudemd__initial_load ◇ CLAUDE.md and rules are now cached per project root rather than once per process.never mined ·never mined never read tengu_cleanup_throttle_marker ◇ If the cleanup marker file is a directory or symlink loop, it's deleted and rebuilt.never mined ·never mined never read tengu_client_data_cache_key ◇ An analytics event, not a feature flag. Claude Code logs it while refreshing its cached account client data, with booleans slot_hit (a cache slot exists for the current account key), slot_changed, legacy_fallback (no slot but the older single clientDataCache exists) and slot_stale.never mined ·never mined never read tengu_client_data_stale_refetch ◇ Claude can refetch startup config when the server says it is stale, but the gate is off.never mined ·never mined never read tengu_cloud_plugins_admission ◇ Cloud plugin admission clears a set-aside session as soon as the decision lands.never mined ·never mined never read tengu_cloud_plugins_consent ◇ You now decide separately whether cloud sessions may use your plugins and your local settings.never mined ·never mined never read tengu_cloud_plugins_late_install ◇ You now decide separately whether cloud sessions may use your plugins and your local settings.never mined ·never mined never read tengu_cloud_plugins_mixed_patch ◇ A refused cloud plugin change no longer throws away the other settings sent with it.never mined ·never mined never read tengu_cloud_plugins_skipped ◇ Plugin forwarding to a cloud session now reports its state and why a run was skipped.never mined ·never mined never read tengu_cobalt_frost ◇ No description from anybody.never mined ·never mined never read tengu_cobalt_heron ◇ Default false. Makes the context tip opus-on-pro-near-limit eligible: on a Pro plan, with an Opus model, after more than half of the usage limit is used, the tip suggests /model and says Sonnet 5 uses the weekly limit more slowly than Opus.v2.1.191 → v2.1.211 ○gone AND v2.1.308 tengu_cobalt_plinth_comfrey ◇ A new flag affects relay routing when reading published artifact files.v2.1.261 → v2.1.309 ◆in the build never read tengu_cobalt_plinth_hornbeam ◇ No description from anybody.v2.1.292 → v2.1.309 ◆in the build never read tengu_cobalt_plinth_laurel ◇ Default false. When on, the Artifact tool's input schema gains an optional lang parameter, a BCP-47 language tag (such as ja or pt-BR) that becomes the published page's <html lang>.v2.1.212 → v2.1.270 ○gone AND v2.1.269 tengu_cobalt_plinth_osier ◇ Default false. When on, listing artifacts or the instances of an Artifact type with any scope other than mine is refused with shared-scope listing is disabled (reason scope_disabled).v2.1.210 → v2.1.267 ○gone AND v2.1.269 tengu_cobalt_wren ◇ No description from anybody.v2.1.138 or earlier → v2.1.309 ◆in the build AND v2.1.308 tengu_code_review_routed ◇ Code review logging now records why a given effort level was picked for your run.never mined ·never mined never read tengu_compact ◇ Automatic conversation compaction now flags when foreign or teammate content is being summarizednever mined ·never mined never read tengu_compact_line_prefix_killswitch ◇ No description from anybody.never mined ·never mined never read tengu_compact_preserved_unanchored ◇ Transcript saving and cleanup logic is consolidated in one place; nothing changes for you.never mined ·never mined never read tengu_compact_replaced_by_hook ◇ Compaction now records when a hook fully replaced Claude Code's own summarization logicnever mined ·never mined never read tengu_compact_streaming_retry ◇ No description from anybody.never mined ·never mined never read tengu_compass_dial ◇ Default false. When true, browsing prompt history shows a dim position label at the top of the prompt box, History N/M (or History while the total is not yet known), and Claude Code counts the history entries to fill in the total.v2.1.157 → v2.1.177 ○gone AND v2.1.308 tengu_compiled_acorn ◇ A new flag controls how poll and mid-turn event text is formatted.never mined ·never mined AND v2.1.308 tengu_concurrent_onquery_detected ◇ An analytics event, not a feature flag. Claude Code logs it, with no fields, when a new query is submitted while another query is still running; the new query does not run, and its user prompts (other than some internal and notification messages) are put back on the input queue.never mined ·never mined never read tengu_concurrent_sessions ◇ An analytics event, not a feature flag. Claude Code logs it at startup, after registering the session and counting the live sessions in its session registry, when that count is 2 or more; it carries num_sessions.never mined ·never mined never read tengu_config_auth_loss_prevented ◇ A second global-config writer goes through the storage backend with better corruption handling.never mined ·never mined never read tengu_config_auto_repaired ◇ An analytics event, not a feature flag. Claude Code logs it when a locked save of the global config re-reads the config file, hits a parse error, and rewrites the file from its cached config instead. It carries file_size_before (the size of the unparseable file) and had_cached_auth (whether the cached config had an oauthAccount or hasCompletedOnboarding: true).never mined ·never mined never read tengu_config_changed ◇ Plugin-provided settings now appear in /config, with their values redacted in telemetrynever mined ·never mined never read tengu_config_fallback_skipped_contention ◇ Saving or deleting config now retries with backoff when the lock is held instead of failing outright.never mined ·never mined never read tengu_config_lock_contention ◇ The settings lock contention counter stops counting unrelated storage failures.never mined ·never mined never read tengu_config_parse_error ◇ Keychain lookups fail quietly instead of crashing, and unparseable config files now get reported.never mined ·never mined never read tengu_config_stale_write ◇ A second global-config writer goes through the storage backend with better corruption handling.never mined ·never mined never read tengu_content_block_healed ◇ A malformed response block is repaired or dropped rather than crashing your session.never mined ·never mined never read tengu_context_tip_classifier_outcome ◇ An analytics event, not a feature flag. Claude Code logs it each time the context tips classifier (a model call with query source context_tip_classifier) finishes, with outcome one of tip, tip_ineligible, no_tip, parse_failure, error or outer_error. Other fields, when present: classifierLogId, durationMs, eligibleIds, featureId, inputTokens, cachedInputTokens, outputTokens and error details.never mined ·never mined never read tengu_continue ◇ claude -p --continue now says a matching session is running in the background instead of silently finding nothing.never mined ·never mined never read tengu_conversation_forked ◇ An analytics event, not a feature flag. Claude Code logs it after it forks the current conversation into a new branch session and saves that session's title (by default the first prompt followed by (Branch) or (Branch N)). It carries message_count and has_custom_title.never mined ·never mined never read tengu_conversation_rewind ◇ Compaction cleanup and rewind both notify a shared hook with a reason string.never mined ·never mined never read tengu_convolute_arcades_retry ◇ A hidden, server-controlled fallback path can now retry refused model turns under different, less visible telemetry namesnever mined ·never mined never read tengu_convolute_arcades_tools ◇ A hidden, server-controlled fallback path can now retry refused model turns under different, less visible telemetry namesnever mined ·never mined never read tengu_coordinator_panel ◇ No description from anybody.v2.1.157 → v2.1.309 ◆in the build never read tengu_copper_gazette ◇ No description from anybody.v2.1.292 → v2.1.309 ◆in the build AND v2.1.308 tengu_copper_kestrel ◇ Default true. Gates effort changes over Remote Control: when on, a connected client can set effortLevel through apply_flag_settings and the session reports its effort to the bridge as effort_level metadata. When off, such a request fails with effort changes over Remote Control are turned off.v2.1.234 → v2.1.309 ◆in the build never read tengu_copper_lantern ◇ Default false. A recall switch for the Claude Code background daemon when it runs as an installed service: the supervisor checks it on each poll, and when it is on it logs service recall flag set, stops its workers, shuts down with cause service_recall and uninstalls the service. A daemon started as a transient process ignores it.v2.1.138 or earlier → v2.1.309 ◆in the build never read tengu_coral_beacon ◇ Default false. When on, the rate-limit options menu adds an option Upgrade to Team plan (labelled Switch to Team plan on Max 20x) for users not on a Team or Enterprise plan, if the upgrade action is enabled. It also lets the rate-limit upsell menu be offered to such users when the upgrade command is available (not disabled by DISABLE_UPGRADE_COMMAND and not an Enterprise plan), in cases where the plan's upgrade paths would otherwise not qualify.v2.1.138 or earlier → v2.1.258 ○gone AND v2.1.251 tengu_coral_fern ◇ Default false. When on, the auto memory prompt gets a ## Searching past context section telling the model to search the memory directory's .md topic files first and session transcript .jsonl logs as a last resort, with an example grep -rn command or Grep tool call for each, and to use narrow search terms.v2.1.138 or earlier → v2.1.146 ○gone AND v2.1.308 tengu_coral_whistle ◇ No description from anybody.never mined ·never mined never read tengu_cosmic_crayon ◇ No description from anybody.v2.1.299 ○gone AND v2.1.308 tengu_cosmic_lake ◇ Default false. When on, the rate-limit-warning notification (the warning shown as a usage limit gets close) is skipped when the limit status is allowed_warning and the limit type is five_hour. Warnings for other limit types still show.v2.1.282 → v2.1.309 ◆in the build never read tengu_cosmic_shore ◇ Default true. When on, a shell command running in the background is stopped once it passes a deadline, capped at 2 hours (or BASH_MAX_TIMEOUT_MS if larger): a command started with run_in_background uses its timeout if given, and otherwise, or when a command is moved to the background later, the deadline is 30 minutes (or BASH_DEFAULT_TIMEOUT_MS if larger); the message telling Claude a command moved to the background can then say when it will be stopped. When off, or when the background deadline is disabled for the session, backgrounded commands get no such deadline, and commands started by Monitor never get it.v2.1.285 → v2.1.309 ◆in the build AND v2.1.308 tengu_cost_threshold_acknowledged ◇ The spend warning and resume prompt now queue with other dialogs instead of popping over your work.never mined ·never mined never read tengu_cozy_dusk ◇ A new numeric remote-config gate was added with clamped bounds.v2.1.261 → v2.1.309 ◆in the build never read tengu_crimson_vector ◇ Default false. When on, and requests go to Anthropic's first-party API (the firstParty provider with no ANTHROPIC_BASE_URL or an Anthropic one), the default maximum base64 size for an image sent to the model rises from 5242880 bytes (5 MB) to 10485760 bytes (10 MB). A model whose own image limits set a maximum size keeps that value.v2.1.138 or earlier → v2.1.309 ◆in the build AND v2.1.308 tengu_cross_session_inbound_changed ◇ A settings row for handling messages from your other sessions exists but is hidden here.never mined ·never mined never read tengu_curried_trinket ◇ No description from anybody.v2.1.290 → v2.1.309 ◆in the build AND v2.1.308 tengu_daemon_control ◇ An analytics event, not a feature flag. Claude Code logs it from the claude daemon start, restart, stop and uninstall subcommands with the outcome: op_start, op_restart, op_stop or op_uninstall plus ok, and depending on the path launcher, reachable, reaped or holderUnverified.v2.1.138 or earlier → v2.1.241 ○gone never read tengu_daemon_install ◇ An analytics event, not a feature flag. Claude Code logs it from claude daemon install with ok: on a refusal it names the reason (disabled, launcher, foreground, or holderAlive with holderUnverified or holderUnknownOrigin), a failed service install logs ok: false alone, and a successful install records whether the daemon was reachable afterwards. claude daemon start and restart also log it with disabled: true when those subcommands are disabled.v2.1.138 or earlier → v2.1.241 ○gone never read tengu_daemon_refuse_stale_upgrade ◇ Default true. When the background daemon's supervisor sees that the Claude Code binary it runs from was replaced by an older build (judged by the build timestamps in the two build names), it refuses the self-restart it would otherwise do to pick up the new binary, keeps the running build and logs refusing self-restart for upgrade. When off, that check is skipped and the upgrade restart goes on through the usual checks.v2.1.203 → v2.1.309 ◆in the build never read tengu_daemon_stale_lock_replaced ◇ Replacing a stale daemon lock now reports the proof type, holder age and whether it ran a different version.never mined ·never mined never read tengu_daemon_startup_crash ◇ The background daemon now restores Cloud gateway sign-in early at startupv2.1.138 or earlier → v2.1.304 ○gone never read tengu_daemon_upgrade_deferred_busy ◇ The background service waits for your turn to finish before restarting for an upgrade.never mined ·never mined never read tengu_daemon_upgrade_respawn_unreachable ◇ An analytics event, not a feature flag. Claude Code logs it when the daemon detects an upgrade and re-spawns itself, and the successor process spawns but does not become reachable in time. The field stderr_captured says whether any of the successor's stderr was captured.v2.1.203 → v2.1.241 ○gone never read tengu_dash_flame ◇ No description from anybody.v2.1.219 → v2.1.309 ◆in the build AND v2.1.308 tengu_dead_probe_ ◇ Temporary probes that checked whether old code paths still ran have been deleted; behaviour is unchanged.never mined ·never mined never read tengu_dead_probe_adopt_ticks_token ◇ One-off counters now show which old code paths are still being used, so they can be removed.never mined ·never mined never read tengu_dead_probe_autoupdater_status ◇ Very old config files with legacy updater values now migrate to an unknown install method.never mined ·never mined never read tengu_dead_probe_bg_legacy_op ◇ Each older background-session operation reports once, showing which legacy paths are still used.never mined ·never mined never read tengu_dead_probe_daemon_origin_auto ◇ An analytics event, not a feature flag. Claude Code logs it, at most once per site per process, when it meets the daemon origin value auto: when parsing --origin auto (which is treated as transient), when formatting the daemon status, and in the daemon status output. The field site names which of the three it was.v2.1.208 → v2.1.220 ○gone never read tengu_dead_probe_enable_auto_mode_flag ◇ Five more leftover diagnostic probes were deleted, with no change to what you see.never mined ·never mined never read tengu_dead_probe_hook_updated_mcp_tool_output ◇ Five more leftover diagnostic probes were deleted, with no change to what you see.never mined ·never mined never read tengu_dead_probe_include_coauthored_by ◇ One-off counters now show which old code paths are still being used, so they can be removed.never mined ·never mined never read tengu_dead_probe_install_counts_cleanup ◇ Five more leftover diagnostic probes were deleted, with no change to what you see.never mined ·never mined never read tengu_dead_probe_iterm2_crash_restore ◇ Two new counters cover swept placeholder bridge sessions and iTerm2 crash-restore probes.never mined ·never mined never read tengu_dead_probe_mcp_subsec_timeout ◇ Most of the one-shot probes checking whether old code paths still run were removed.never mined ·never mined never read tengu_dead_probe_pinned_sidecar ◇ Most of the one-shot probes checking whether old code paths still run were removed.never mined ·never mined never read tengu_dead_probe_plugin_toplevel_experimental ◇ Most of the one-shot probes checking whether old code paths still run were removed.never mined ·never mined never read tengu_dead_probe_preserved_segment ◇ Temporary probes that checked whether old code paths still ran have been deleted; behaviour is unchanged.never mined ·never mined never read tengu_dead_probe_speculation_overlay ◇ Five more leftover diagnostic probes were deleted, with no change to what you see.never mined ·never mined never read tengu_dead_probe_taskstop_shell_id ◇ Five more leftover diagnostic probes were deleted, with no change to what you see.never mined ·never mined never read tengu_dead_probe_tool_alias_exec ◇ Five more leftover diagnostic probes were deleted, with no change to what you see.never mined ·never mined never read tengu_dead_probe_voice_enabled_flat ◇ Five more leftover diagnostic probes were deleted, with no change to what you see.never mined ·never mined never read tengu_declared_tool_set_held ◇ A new event tracks how often a session has to defer late tools or redeclare dropped ones.never mined ·never mined never read tengu_declared_tools_dropped_at_compaction ◇ Claude Code can now drop unused MCP tools from a session's declared-tools list when compacting the conversationnever mined ·never mined never read tengu_deep_ocean_current ◇ No description from anybody.never mined ·never mined never read tengu_defer_cap_ms ◇ Backgrounding a task now reads the terminal title and agent from shared stores.v2.1.195 → v2.1.309 ◆in the build never read tengu_deferred_stub_tool ◇ Default true, read once per session and then kept. When on and deferred tool loading is in use, a request's tool list gains a tool named DeferredToolPlaceholder with defer_loading set, described as Reserved placeholder that keeps deferred tool loading active; never call this tool., unless one is already there. It is never added under HIPAA restrictions, or while CLAUDE_CODE_DISABLE_EXPERIMENTAL_BETAS is set and tool search is not forced on.v2.1.207 → v2.1.309 ◆in the build AND v2.1.308 tengu_deferred_tools_pool_change ◇ Deferred-tool telemetry now tracks resurfaced tools and whether tool search is unavailablenever mined ·never mined never read tengu_device_bash_output_limit ◇ Device bash limit and timeout events dropped their session-matching field.never mined ·never mined never read