Feature flags, all 2,081.

tengu_haze_glass ◇ When only one org memory store is found, it is no longer picked for you automatically.v2.1.210 → v2.1.308 ◆in the build AND v2.1.281 tengu_hazel_osprey ◇ Default false. Turns on the context-hint-2026-04-09 beta for main-thread requests that carry first-party betas: once an estimate of savable tokens reaches 20000, the request body gets context_hint with enabled: true and a target_tokens_saved from tengu_hazel_osprey_floor (default 75000, omitted unless above 0). API errors such as 422, 424, 409, 529 or an invalid request turn the hint off again.v2.1.138 or earlier → v2.1.308 ◆in the build AND v2.1.110 tengu_heron_tallow ◇ Default false. Changes the hint added when the model calls a tool that is not available: for a known tool it reads <tool> is disabled for this session, in subagents as well as here. instead of exists but is not enabled in this context, and for Glob or Grep when Bash is also unavailable it reads <tool> is disabled for this session. CLAUDE_CODE_HERON_TALLOW turns on the same wording.v2.1.217 → v2.1.228 ○gone AND v2.1.217 tengu_herring_clock ◇ Default false. When auto memory is enabled, turns on team memory: a team directory inside the auto-memory directory is loaded into the memory prompt next to the personal one, the team scope can be synced at startup, and writes to it are scanned for secrets (cannot be written to team memory). A non-empty CLAUDE_MEMORY_STORES turns team memory on as well.v2.1.138 or earlier → v2.1.209 ○gone AND v2.1.210 tengu_hover_rest ◇ The new storage layer now touches nearly every part of the product but is never actually created.v2.1.221 → v2.1.308 ◆in the build AND v2.1.304 tengu_idle_amber_finch ◇ Default false. When on, Claude Code stops offering an upgrade to Max: the usage-limit menu drops its Upgrade to Max option, the shared upgrade-availability check returns false (so, for example, the five-hour limit message no longer suggests /upgrade), and the startup subscription check no longer reports an existing Claude Max subscription.v2.1.138 or earlier → v2.1.308 ◆in the build AND v2.1.308 tengu_iridescent_crystal ◇ Auto-mode's safety classifier gained a new 'arbiter' mode that can skip some server checksv2.1.269 → v2.1.308 ◆in the build AND v2.1.281 tengu_iridescent_pascal ◇ New gated flag tengu_iridescent_pascal can skip redundant permission-context resetsv2.1.280 → v2.1.291 ○gone AND v2.1.292 tengu_iterative_lightning ◇ When a session is resumed, Claude can now be warned about tool calls that were lost, but only when this is switched on remotelyv2.1.299 → v2.1.308 ◆in the build AND v2.1.302 tengu_jade_anvil_4 ◇ The usage-limit dialog now takes the reset time for its auto-continue option from a snapshot made when the dialog opensv2.1.138 or earlier → v2.1.308 ◆in the build AND v2.1.296 tengu_jaunty_umbrella ◇ A new check covers one specific tool in a permission mode named arbiter and stays off unless the server turns it onv2.1.300 → v2.1.308 ◆in the build AND v2.1.300 tengu_jazzy_goblet ◇ Changing the automatic model-fallback-on-refusal setting now fires its own telemetry event and saves via userSettingsv2.1.269 → v2.1.308 ◆in the build AND v2.1.269 tengu_juniper_relay ◇ Saved feedback drafts can be reviewed and sent over the SDK, but the surface is off by default.v2.1.212 → v2.1.308 ◆in the build AND v2.1.251 tengu_lantern_spool ◇ A server flag can widen when subagent requests send the extended usage-limit header.v2.1.176 → v2.1.308 ◆in the build AND v2.1.292 tengu_lantern_wick ◇ Default false. When on, and the usage limit's grace window has just become active, the next main-thread follow-up request in a turn (not the first request of a user turn, and not one resuming incomplete thinking) gets a hidden message saying the usage limit is reached and the grace window is active, ending Wrap up: finish or checkpoint; don't start subagents or long work. A new user turn clears the pending notice without adding it.v2.1.217 → v2.1.226 ○gone AND v2.1.308 tengu_larch_pavise ◇ A per-session workshop document gets snapshotted with your session files and is only writable in plan mode.v2.1.219 → v2.1.267 ○gone AND v2.1.219 tengu_lilac_dune ◇ A new 'summary' hint for thinking blocks exists in the code but is gated off for nowv2.1.269 → v2.1.281 ○gone AND v2.1.282 tengu_lively_beaver ◇ A full account-on-hold sign-in flow was added but stays unreachable pending a server-side gatev2.1.248 → v2.1.301 ○gone AND v2.1.235 tengu_loggia_carousel ◇ Default false. Turns on the model refusal fallback: messages with subtype model_refusal_fallback are rendered, the /config toggle Switch models when a message is flagged (setting switchModelsOnFlag, default on) is shown, and with that setting on and a first-party API a fallback model can be requested from the server. CLAUDE_CODE_DISABLE_REFUSAL_FALLBACK forces it off; the flag's value is also sent to the VS Code extension in its experiment_gates notification.v2.1.152 → v2.1.169 ○gone AND v2.1.308 tengu_loud_sugary_rock2 ◇ Default false. Applies only when the main model is claude-opus-4-7. When on, the system prompt gets a # Thinking system reminder section, and each regular user prompt after the first assistant reply gets a meta reminder: Respond with just the action or changes and without a thinking block, unless this is a redesign or requires fresh reasoning. The reminder is not added with a custom system prompt or with thinking disabled, and a cached client-data value of "true" or "false" overrides the flag.v2.1.138 or earlier → v2.1.152 ○gone AND v2.1.308 tengu_lovely_umbrella ◇ Policy-limits change notifications can now be skipped under an experiment flagnever mined ·never mined AND v2.1.281 tengu_lucky_brooks ◇ When switched on remotely, a first message over 2000 characters is cut to its first 1500 before Claude Code generates a session title from itv2.1.282 → v2.1.308 ◆in the build AND v2.1.282 tengu_lucky_garden ◇ A reworked way of capturing your shell's functions, aliases and options for Bash commands is in the code, off unless switched on remotelyv2.1.287 → v2.1.289 ○gone AND v2.1.287 tengu_madrone_spindle_corbel ◇ Publish results can note when an artifact was seeded from a comment thread, prompting a reply.v2.1.260 → v2.1.308 ◆in the build AND v2.1.260 tengu_magical_pixel ◇ New tip promotes Claude Code Desktop's per-session iOS Simulator pane for iOS app workv2.1.265 → v2.1.308 ◆in the build AND v2.1.281 tengu_magical_pixel_startup ◇ The macOS iOS-app intent check now always reads the tengu_magical_pixel gate, and tengu_magical_pixel_startup is no longer usedv2.1.278 ○gone AND v2.1.281 tengu_malformed_tool_use_clean_retry ◇ Default false. When a response stops with tool_use but no tool call could be parsed, Claude Code retries once; with this flag on, the malformed assistant messages are tombstoned and dropped from the retry and the nudge reads The previous response failed to produce a valid tool call. Please retry the tool call now. With it off, the malformed messages stay in the history and the nudge is Your tool call was malformed and could not be parsed. Please retry.v2.1.174 → v2.1.250 ○gone AND v2.1.308 tengu_managed_config_prefetch_off ◇ Claude Code now fetches organization-managed settings early at startup by default when it connects to Anthropic directly, unless switched off remotelyv2.1.293 → v2.1.308 ◆in the build AND v2.1.293 tengu_maple_pier ◇ The auto-mode suggestion would only appear in sessions where auto mode can actually be used.v2.1.152 → v2.1.308 ◆in the build AND v2.1.285 tengu_maple_sundial ◇ Default false. Turns on the reworked /config panel: some setting labels change (for example Verbose in place of Verbose output), rows such as Notifications and Agents view are added, and the setting list is regrouped. In interactive sessions /vim and /output-style become hidden commands that say they moved to /config, and the standalone /output-style command is disabled.v2.1.163 → v2.1.308 ◆in the build AND v2.1.163 tengu_maple_tide ◇ Default false. When true, enables the /goal command, which sets a condition for Claude to keep working toward, in interactive sessions and, in its non-interactive variant, in non-interactive or remote sessions; it also restores a goal from the transcript when a session is resumed.v2.1.138 or earlier ○gone AND v2.1.308 tengu_marble_lark ◇ Default false. When auto memory is enabled, turns on personal memory sync: the startup sync also pulls and pushes the user memory scope (still subject to the allow_memory_sync policy), and writes into the auto-memory directory are scanned for secrets with the error Memory is synced to your account. When it reads false, an in-flight user scope push stops with personal sync disabled.v2.1.147 → v2.1.209 ○gone AND v2.1.308 tengu_marl_cormorant ◇ Default false. When on, the Bash tool description gains the line Command output is displayed to you, not reliably to the user. The line is also added when CLAUDE_CODE_MARL_CORMORANT is set to a true value, when the cached client data has tengu_marl_cormorant: true, or when the model has the opus_5_prompt_bundle capability and tengu_fennel_godwit is off.never mined ·never mined AND v2.1.215 tengu_mcp_memory_cgroup ◇ On Linux you can force MCP servers into their own memory cgroup with an environment variable.v2.1.239 → v2.1.245 ○gone AND v2.1.246 tengu_mcp_retry_failed_remote ◇ Default true. When true, remote MCP servers (http, sse, claudeai-proxy) whose initial connect failed transiently (HTTP 500, 502, 503, 504, network errors such as ECONNREFUSED or ETIMEDOUT, or an sse failure with no error code) are retried. One connect path retries each such server up to 3 times with backoff starting at 1000 ms and capped at 30000 ms; the other retries the still-failed servers after 500, 1500 and 4000 ms.v2.1.138 or earlier ○gone AND v2.1.308 tengu_mcp_subagent_prompt ◇ Default false. Chooses the instructions Claude gets when an MCP tool result is too large and is saved to a file. Off gives the generic text (read it in chunks with offset and limit, then state how much was read); on gives format-specific steps (probe JSON with jq, slice long lines by character range, or read a stated number of lines per chunk) and suggests doing the reading in a subagent. MCP_TRUNCATION_PROMPT_OVERRIDE overrides it: legacy forces the generic text, any other non-empty value the new one.v2.1.138 or earlier → v2.1.308 ◆in the build AND v2.1.308 tengu_memoized_owl ◇ The fixed browser setup reply in the desktop app and on claude.ai is now controlled by a newly named switch, so any earlier rollout starts overv2.1.301 → v2.1.308 ◆in the build AND v2.1.301 tengu_merry_tide ◇ One tool can now be offered when an extra condition holds and a remote switch is on, and the shared tool description needs that same conditionv2.1.299 → v2.1.308 ◆in the build AND v2.1.299 tengu_mill_orange ◇ Chinese, Japanese and Korean messages without spaces could trigger memory recall, but that branch is off here.v2.1.219 → v2.1.308 ◆in the build AND v2.1.221 tengu_mint_lanes ◇ Default false. Turns on kind lanes in the slash command menu, as does CLAUDE_CODE_ENABLE_MENU_KIND_LANES: each command row gets a kind label (info commands show as config, action shows nothing) and a [source] tag, and uses the command's menuDescription when it has one.v2.1.163 → v2.1.308 ◆in the build AND v2.1.163 tengu_misty_anchor ◇ Cloud sessions can fetch a server-built memory block into the system prompt, off in this build.v2.1.246 → v2.1.308 ◆in the build AND v2.1.271 tengu_moss_anchor ◇ The built-in fallback for starting sessions in auto mode switched from off to on, so it applies unless the server turns it offv2.1.152 → v2.1.308 ◆in the build AND v2.1.302 tengu_mossy_wirth ◇ Reading project files can return an uploaded PDF or Word file itself, and writes can check a version first; both are off unless switched on remotelyv2.1.299 → v2.1.308 ◆in the build AND v2.1.299 tengu_moth_copse ◇ Default false. When true, Claude Code takes the memory stores code paths that a non-empty CLAUDE_MEMORY_STORES environment variable turns on; for example, AutoMem memory entries are filtered out of the loaded memory files. One check differs: only the environment variable, not this flag, makes files the memory mode when no earlier choice applies.v2.1.138 or earlier → v2.1.308 ◆in the build AND v2.1.217 tengu_noreread_q7m_velvet ◇ No description from anybody.never mined ·never mined AND v2.1.91 tengu_ochre_bittern ◇ A helper that always returned false now checks a remote gate first.v2.1.260 → v2.1.308 ◆in the build AND v2.1.260 tengu_ochre_dunlin ◇ When staged files fail to land, Claude Code logs a warning and can tell the model the files may be missing or out of datev2.1.295 → v2.1.308 ◆in the build AND v2.1.300 tengu_ochre_hollow ◇ Artifact syntax highlighting and mermaid diagrams are now always on; three other gates were renamed and stay off.v2.1.138 or earlier → v2.1.238 ○gone AND v2.1.239 tengu_ochre_kite ◇ Claude Code now recovers gracefully when a safety filter cuts off a response mid-tool-callv2.1.273 → v2.1.281 ○gone AND v2.1.282 tengu_ochre_plover ◇ Claude can auto-draft replies to artifact comments, but it's off by default and declines by default.v2.1.260 → v2.1.308 ◆in the build AND v2.1.260 tengu_ochre_wren ◇ An alternate opening line for Claude's system prompt can be switched on with an environment variable.v2.1.239 → v2.1.294 ○gone AND v2.1.248 tengu_parchment_fern ◇ An alternate prompt wording for two tools exists but stays off unless you set an environment variable.v2.1.248 → v2.1.283 ○gone AND v2.1.223 tengu_passport_quail ◇ Memory extraction can now turn itself on when post-turn-memory config is present, bypassing its usual feature gatev2.1.138 or earlier → v2.1.308 ◆in the build AND v2.1.305 tengu_plugin_binary_assets ◇ Default false. When true, or when the CLAUDE_CODE_PLUGIN_BINARY_ASSETS env var is 1, true, yes or on, Claude Code provisions the native binaries a plugin declares under binaries in its .claude-plugin/plugin.json into the plugin's bin directory, keyed to their declared sha256 values. Provisioning also requires the plugin's marketplace to pass a separate check, and a plugin root that is not a real directory is refused.v2.1.196 → v2.1.308 ◆in the build AND v2.1.196 tengu_post_compact_survey ◇ Default false. When true, Claude Code may show a feedback survey (survey_type: "post_compact") after a compaction: once a user or assistant message follows the newest compaction marker, it opens with probability 0.2. It stays hidden when another survey is open, product feedback is disallowed by policy, or CLAUDE_CODE_DISABLE_FEEDBACK_SURVEY is set.v2.1.138 or earlier → v2.1.308 ◆in the build AND v2.1.308 tengu_quiet_ember ◇ Responsive mode is no longer offered when Claude Code runs inside Slack or Teamsnever mined ·never mined AND v2.1.281 tengu_quiet_harbor ◇ Default false. Supplies the background daemon cold-start mode when neither CLAUDE_CODE_DAEMON_COLD_START (transient or ask) nor the daemonColdStart setting is set: true gives ask, false gives transient. With ask, when no background daemon is running and the install prompt has not been dismissed, Claude Code (under further conditions) returns an install request telling you to run claude daemon install instead of spawning a transient daemon.v2.1.138 or earlier → v2.1.308 ◆in the build AND v2.1.120 tengu_quiet_hopcroft ◇ When a hook cannot check a call, Claude Code now blocks the call instead of letting it throughnever mined ·never mined AND v2.1.288 tengu_quiet_slate_wren ◇ Default false. When on, Claude Code no longer shows the Use your existing Claude Pro plan with Claude Code · /login to activate notice to someone who is not signed in with a subscription but whose claude.ai account has Claude Pro. The equivalent notice for Max accounts is not affected.v2.1.138 or earlier → v2.1.308 ◆in the build AND v2.1.308 tengu_quizzical_shore ◇ A hidden hint can appear on thinking blocks per model, off by default.v2.1.260 → v2.1.281 ○gone AND v2.1.282 tengu_radiant_fox ◇ A notice for Max and Team plans about an API credit link is built in, with its wording supplied remotely and a limit on how often it showsnever mined ·never mined AND v2.1.302 tengu_radiant_hare ◇ A new remote switch can stop accepted edits from skipping the auto mode safety check; it is off by defaultv2.1.305 → v2.1.308 ◆in the build AND v2.1.305 tengu_radiant_heron ◇ A Remote Control background process can now tell cloud sessions they may use your local folder and its tools, if a server switch and several checks allow itv2.1.283 → v2.1.308 ◆in the build AND v2.1.283 tengu_reactive_compact_remote ◇ Default false. Read only when CLAUDE_CODE_REMOTE is set, and latched for the rest of the process on first read. In remote sessions with the flag off, reactive compaction after a context-limit error and precomputed compaction are disabled, and proactive auto-compaction is no longer held back for reactive handling or routed through it. Outside remote sessions the flag is not consulted and these paths behave as if it were on.v2.1.172 → v2.1.308 ◆in the build AND v2.1.308 tengu_reactive_zephyr ◇ In served sessions, Claude can no longer edit settings files, and permission prompts for network paths are never approved automaticallynever mined ·never mined AND v2.1.288 tengu_read_dedup_killswitch ◇ The Read tool accepts allow_large, letting Claude read an oversized text file in one go when there is still room in the conversationv2.1.138 or earlier → v2.1.308 ◆in the build AND v2.1.296 tengu_remote_auto_mode_include_destructive_mcp ◇ Destructive MCP tools could stop skipping permission prompts in remote auto mode, but that is off.v2.1.222 → v2.1.308 ◆in the build AND v2.1.259 tengu_review_workflow_routing ◇ Two server-side switches that steered review behaviour are gone, so reviews no longer vary by remote config.v2.1.160 → v2.1.231 ○gone AND v2.1.232 tengu_russet_plover ◇ Artifact syntax highlighting and mermaid diagrams are now always on; three other gates were renamed and stay off.v2.1.239 → v2.1.250 ○gone AND v2.1.240 tengu_rustling_orbit ◇ A missing policy-limits response with no cache no longer counts as a failure; the session runs with no restrictions instead.v2.1.259 → v2.1.308 ◆in the build AND v2.1.304 tengu_salt_marsh ◇ Memory citations can now be tagged inline with markupv2.1.205 → v2.1.308 ◆in the build AND v2.1.275 tengu_satchel_banjo ◇ The dialog for buying extra usage now reads an offer, a remaining monthly bundle cap and custom discount tiers sent from the serverv2.1.217 → v2.1.308 ◆in the build AND v2.1.304 tengu_scalable_quiche ◇ Artifacts could keep being watched even when your session is not running, but that is off.v2.1.227 → v2.1.238 ○gone AND v2.1.239 tengu_scratch ◇ Claude may soon be told to use a per-session scratchpad directory for temp files instead of /tmp.v2.1.138 or earlier → v2.1.308 ◆in the build AND v2.1.252 tengu_sepia_moth ◇ A precompute-compaction toggle is gated off by default and has no effect even if saved.v2.1.138 or earlier → v2.1.308 ◆in the build AND v2.1.300 tengu_sequential_summit ◇ No description from anybody.v2.1.301 → v2.1.303 ○gone AND v2.1.308 tengu_session_name_uniqueness ◇ Duplicate session names on one machine now get a suffix automatically, and callers are told the new name.v2.1.232 → v2.1.308 ◆in the build AND v2.1.232 tengu_shale_finch ◇ Default false. When on, subagents that are not teammates and are not spawned with an exact tool list lose the TodoWrite, TaskCreate, TaskGet, TaskUpdate and TaskList tools.v2.1.138 or earlier → v2.1.285 ○gone AND v2.1.308 tengu_sharded_snowflake ◇ claude plugin eval is now generally available, with a server-side switch that can turn it off per sessionv2.1.269 → v2.1.308 ◆in the build AND v2.1.269 tengu_shimmying_stardust ◇ Background-hint timing for sync subagents now uses its own separate, oddly-named gatev2.1.273 → v2.1.308 ◆in the build AND v2.1.281 tengu_shining_fractals ◇ A queued-state enum gains a member for queued MCP elicitation requests.v2.1.145 → v2.1.308 ◆in the build AND v2.1.284 tengu_silent_harbor ◇ Claude Code labels and groups each part of its system prompt, and a new detached mode skips a setup step; the prompt text is the samev2.1.176 → v2.1.308 ◆in the build AND v2.1.301 tengu_silk_almanac ◇ No description from anybody.v2.1.187 → v2.1.204 ○gone AND v2.1.187 tengu_silk_hinge ◇ The message-timestamps toggle in settings is still gated and not generally shown.v2.1.138 or earlier → v2.1.308 ◆in the build AND v2.1.288 tengu_skills_dashboard_enabled ◇ The skills dashboard drops entries with unrecognised health values, but the dashboard itself stays off.v2.1.140 → v2.1.308 ◆in the build AND v2.1.223 tengu_slate_finch ◇ Default false. When on and the account's plan is pro, the effort messages (Set effort level to ... and Current effort level: ...) describe high with an added note that it burns fastest and that medium handles most tasks.v2.1.138 or earlier → v2.1.308 ◆in the build AND v2.1.91 tengu_slate_harbor_experiment ◇ Default false. When on, or when CLAUDE_CODE_NEW_INIT is set, /init sends the alternate multi-phase prompt instead of the original one that begins Please analyze this codebase and create a CLAUDE.md file. The /init command description follows CLAUDE_CODE_NEW_INIT alone.v2.1.138 or earlier → v2.1.308 ◆in the build AND v2.1.308 tengu_slate_moth ◇ Default false. Makes the context tip large-context-stale-files eligible: when the context is over 300,000 tokens and more than 100,000 of them are stale file tokens, the tip suggests /compact.v2.1.191 → v2.1.211 ○gone AND v2.1.308 tengu_slate_thimble ◇ Memory extraction can now turn itself on when post-turn-memory config is present, bypassing its usual feature gatev2.1.138 or earlier → v2.1.308 ◆in the build AND v2.1.269 tengu_smooth_chipmunk ◇ Auto mode's dangerous-action classifier now runs server-side and blocks actions it flags or can't reach.v2.1.261 → v2.1.308 ◆in the build AND v2.1.282 tengu_snappy_pelican ◇ Claude can now ask a subagent for lower or higher effort than its own, but the option is off unless switched onv2.1.296 → v2.1.308 ◆in the build AND v2.1.296 tengu_snippet_save ◇ No description from anybody.never mined ·never mined AND v2.1.59 tengu_snug_platypus ◇ Claude Code adds two server-controlled switches, a path for remote-session tools over MCP, and a new label for Claude.ai proxy serversnever mined ·never mined AND v2.1.290 tengu_sorrel_avocet ◇ Claude could start deep research on its own instead of waiting for you to ask, if enabled.v2.1.219 → v2.1.308 ◆in the build AND v2.1.281 tengu_sorrel_trellis_newel ◇ Artifact-comment auto-reply gains a fork/handoff path behind the tengu_sorrel_trellis_newel gatev2.1.268 → v2.1.308 ◆in the build AND v2.1.269 tengu_sorrel_trellis_tenon ◇ Auto-replies to artifacts can now be delivered through an alternate, forked/subagent pathv2.1.269 → v2.1.308 ◆in the build AND v2.1.271 tengu_sorted_flute ◇ A new internal check can now silently skip saving flag settings changes to diskv2.1.280 → v2.1.285 ○gone AND v2.1.286 tengu_sparrow_ledger ◇ Default false. When on, adds a Work step by step: section to the system prompt telling the model to reproduce the symptom before editing, edit the source, then re-observe the symptom until it is gone. The CLAUDE_CODE_VERIFY_PROMPT environment variable turns on the same section.v2.1.138 or earlier → v2.1.205 ○gone AND v2.1.308 tengu_spicy_pearl ◇ No description from anybody.v2.1.299 → v2.1.308 ◆in the build AND v2.1.308 tengu_sprightly_lagoon ◇ Default false. When on, the checks for per-turn effort and per-turn timing both report unsupported, so requests carry no per-turn effort or now values and those checks add no per-turn-control-2026-07-01 beta.v2.1.260 → v2.1.301 ○gone AND v2.1.308