Feature flags, all 2,081.

tengu_feature_ ◇ No description from anybody.v2.1.280 → v2.1.308 ◆in the build never read tengu_feature_bad ◇ Feature-bad/feature-sad telemetry now sanitizes the error_code field before sending itnever mined ·never mined never read tengu_feature_sad ◇ Feature-bad/feature-sad telemetry now sanitizes the error_code field before sending itnever mined ·never mined never read tengu_federated_flask ◇ /autofix-pr can now refuse a pull request that PR Steward is watching or working on, so two agents don't push to it at oncev2.1.284 → v2.1.308 ◆in the build AND v2.1.284 tengu_federated_robin ◇ Default true. When on, a failed fetch of /api/oauth/usage is remembered for the OAuth token that sent it: an auth rejection (401, or 403 with an Anthropic error body) for an hour, and a 429 or other 403 for the Retry-After time (5 minutes when none is given, at most an hour). Until then a fetch with that token throws the remembered error instead of asking the server again. Turning it off clears the remembered errors and records no new ones.v2.1.284 → v2.1.306 ○gone never read tengu_feedback_draft_discarded ◇ Oversized feedback submissions now tell you to drop the transcript or shorten the details.never mined ·never mined never read tengu_feedback_draft_submitted ◇ Feedback now lets you tag a failure mode and task category, cycled with the arrow keys.never mined ·never mined never read tengu_feedback_survey_config ◇ The feedback survey checks for itself whether you're mid-prompt, so it interrupts you less.v2.1.162 → v2.1.308 ◆in the build AND v2.1.247 tengu_feedback_survey_event ◇ MCP notification wiring for VS Code and daemon sessions now goes through one shared path.never mined ·never mined never read tengu_feedback_turnoff_prompt_declined ◇ Toggling Claude-drafted feedback now records whether you changed it from the card or from /config.never mined ·never mined never read tengu_feedback_turnoff_prompt_shown ◇ Toggling Claude-drafted feedback now records whether you changed it from the card or from /config.never mined ·never mined never read tengu_fennel_godwit ◇ On Opus 5, Claude is told not to spawn subagents, workflows or deep research unless you ask.v2.1.219 → v2.1.301 ○gone AND v2.1.219 tengu_fennel_kite_model ◇ Default "". A model-ID substring: when the current model's ID contains it, and the base preconditions hold (the default Anthropic API base URL or CLAUDE_CODE_MID_CONVERSATION_SYSTEM set, plus other provider and model checks), requests send the mid-conversation-system-2026-04-07 beta header and message building switches to its mid-conversation-system handling. A model ID matching CLAUDE_CODE_MID_CONVERSATION_SYSTEM is accepted first, and a non-empty cached client-data value of the same name is used in place of the flag.v2.1.138 or earlier → v2.1.153 ○gone AND v2.1.308 tengu_fern_plover ◇ No description from anybody.v2.1.287 → v2.1.308 ◆in the build never read tengu_fg_left_arrow_agents ◇ Default false. Turns on the left-arrow shortcut to the agents view, which is also on once the global config has hasUsedAgentsFleet set; both paths also require a local (non-remote) workspace and agent view not being disabled by CLAUDE_CODE_DISABLE_AGENT_VIEW or the disableAgentView setting. When active, the footer can show a ← for agents hint, the left arrow key can open the agents view, and the config panel gains a ← opens agents toggle (leftArrowOpensAgents, default true) that turns the hint and the key off.v2.1.138 or earlier → v2.1.153 ○gone AND v2.1.308 tengu_fgts ◇ Tool descriptions can differ by session proactive level; the streaming behaviour behind it is off.v2.1.138 or earlier → v2.1.308 ◆in the build AND v2.1.281 tengu_file_changed ◇ Whole-file rewrites now count removed lines instead of reporting zero.never mined ·never mined never read tengu_file_history_rewind_failed ◇ A new telemetry event logs when a file-history rewind fails entirely.never mined ·never mined never read tengu_filtered_orphaned_thinking ◇ No description from anybody.never mined ·never mined never read tengu_filtered_orphaned_thinking_message ◇ Orphaned-thinking-message telemetry no longer fires duplicate events for the same messagenever mined ·never mined never read tengu_fizzy_bonbon ◇ A new gate affects how fast-mode tool permissions are resolvedv2.1.267 → v2.1.308 ◆in the build never read tengu_fizzy_grove ◇ A gate decides which of two error codes an interrupt-triggered abort reports.v2.1.257 → v2.1.308 ◆in the build AND v2.1.280 tengu_fizzy_minsky ◇ In the desktop app and on claude.ai, a browser request with no extension connected can get a fixed setup reply instead of a model answerv2.1.300 ○gone AND v2.1.301 tengu_fizzy_petal ◇ No description from anybody.v2.1.288 → v2.1.308 ◆in the build AND v2.1.308 tengu_fizzy_treasure ◇ Two new switches set by Anthropic's servers are built in, one falling back to off and one to onv2.1.290 → v2.1.308 ◆in the build AND v2.1.290 tengu_flag ◇ A plan approval can only be answered once, and a mismatched answer no longer hangs.never mined ·never mined never read tengu_fleet_needs_input_nudge ◇ No description from anybody.v2.1.205 → v2.1.209 ○gone AND v2.1.205 tengu_fleet_nudge_state ◇ An analytics event, not a feature flag. Claude Code logs it when it refreshes its list of other sessions and the counts of sessions that need input, are done, or finished successfully differ from the previous refresh. It carries needs_input_count, done_count, succeeded_count, and increased (whether the needs-input count went up).never mined ·never mined never read tengu_fleet_past_sessions ◇ Default false; CLAUDE_CODE_FLEET_PAST_SESSIONS set to true also turns it on. When on, the agent view loads past sessions and lists them as earlier entries, and claude --resume given no session ID, name, fork or other session option (such as --model, --permission-mode or --settings) opens the agent view. The --resume change does not apply while agent view is disabled by CLAUDE_CODE_DISABLE_AGENT_VIEW or the disableAgentView setting.v2.1.203 → v2.1.308 ◆in the build AND v2.1.203 tengu_fleetview_dead_epoch_offer ◇ Reopening a job that died offline now asks before respawning, and remembers when it's gone.never mined ·never mined never read tengu_fleetview_earlier_loaded ◇ An analytics event, not a feature flag. Claude Code logs it when the fleet view finishes loading its earlier sessions and the load returned at least one entry, unless the load was cancelled. It carries count, the number of entries loaded.never mined ·never mined never read tengu_fleetview_empty_state_shown ◇ Picking a group header in Fleetview no longer bounces your selection back and forth.never mined ·never mined never read tengu_fleetview_fold_shown ◇ The press-again-to-delete confirmation is more reliable, and escape now dismisses it properly.never mined ·never mined never read tengu_fleetview_onboarding_v2 ◇ Default false. When true and the fleet view's peak count of concurrently active or blocked sessions is below 3, the fleet view loads suggested agent tasks: on the first-party API, outside essential-traffic-only mode, it lists up to 5 of your recently merged pull requests with gh pr list (anyone's if you have none) and asks a model to fill in each suggestion's scope, falling back to generic scopes, cached for 10 minutes. It also lets the state-grouped list show an empty-bucket hint.v2.1.193 → v2.1.202 ○gone never read tengu_fleetview_peers ◇ Fleet view can show other running sessions it never tracked, once two gates are on.v2.1.224 → v2.1.308 ◆in the build AND v2.1.305 tengu_fleetview_pr_batch ◇ Fleet view still keeps peer session adoption off and batched PR status on by default.v2.1.138 or earlier → v2.1.308 ◆in the build AND v2.1.305 tengu_fleetview_simple ◇ The rule for turning on fleet view's simple mode is unchanged, but the way that code looks up keyboard shortcuts changedv2.1.216 → v2.1.308 ◆in the build AND v2.1.299 tengu_flickering_lagoon ◇ send_message can route messages to a bridge address through a matching MCP tool, behind a remote switch that defaults to offv2.1.299 → v2.1.308 ◆in the build AND v2.1.299 tengu_flickering_rain ◇ Bash, PowerShell, and Monitor tools can now declare extra allowed network domains for sandboxed commandsnever mined ·never mined AND v2.1.271 tengu_flint_harbor ◇ No description from anybody.never mined ·never mined AND v2.1.94 tengu_flint_harbor_prompt ◇ Default {}. Overrides for the /team-onboarding command: a string prompt replaces the built-in prompt template, a string guideTemplate replaces the text inserted at {{GUIDE_TEMPLATE}}, and a number windowDays sets how many days of usage data are scanned, floored and clamped to 1 through 365 (default 30). Fields of the wrong type use the built-in values.v2.1.138 or earlier → v2.1.308 ◆in the build AND v2.1.248 tengu_flint_harbor_share ◇ Default false. Turns on sharing for /team-onboarding: it enables the ShareOnboardingGuide tool, appends instructions for calling that tool to the command's prompt, and lets the team-onboarding-share tip be shown. Each of these also requires the allow_team_onboarding policy, an OAuth access token, and one further check to pass.v2.1.138 or earlier → v2.1.308 ◆in the build AND v2.1.128 tengu_foamy_sedgewick ◇ A new instruction tells Claude to treat downloads and unpacked archives as untrusted and keep them apart from its own scriptsnever mined ·never mined AND v2.1.290 tengu_foamy_spring ◇ A new gate is on by default and stays on even if its reader fails.v2.1.251 → v2.1.308 ◆in the build never read tengu_fold_tool ◇ Default false. In an interactive session, the message list outside the transcript view hides every message whose uuid is listed in the foldedUuids of a fold_boundary system message. Non-interactive sessions treat the flag as off.v2.1.198 → v2.1.209 ○gone AND v2.1.198 tengu_fork_agent_query ◇ Fork-agent query telemetry now records whether a query was eligible for the teleport relaynever mined ·never mined never read tengu_fork_subagent_enabled ◇ Forked subagents are on by default in interactive sessions; set CLAUDE_CODE_FORK_SUBAGENT to false to opt out.never mined ·never mined never read tengu_forked_agent_default_turns_exceeded ◇ Unexpected telemetry values are replaced with a placeholder instead of being sent raw.never mined ·never mined never read tengu_frame_publish_context ◇ Frame publish context is now permanently off and no server setting can enable it.v2.1.196 → v2.1.247 ○gone AND v2.1.248 tengu_fresh_heron ◇ Claude Code has a new remotely controlled switch that counts as on when the server sends no value, and nothing yet shows what it controlsv2.1.289 → v2.1.308 ◆in the build AND v2.1.289 tengu_frontmatter_shadow_unknown_key ◇ An analytics event, not a feature flag. Claude Code logs it when the frontmatter of a skill, agent, or output style has a key that its strict check schema for that kind does not recognize, at most once per surface and key. It carries surface and detail, derived from the kind and the unknown key.never mined ·never mined never read tengu_fullscreen_crash_auto_off ◇ Accepting fullscreen now runs as a trial, and two crashed launches disable it on that machine.never mined ·never mined never read tengu_fullscreen_upsell_trial_persisted ◇ A trouble-free fullscreen trial saves the setting, unless you changed it yourself mid-trial.never mined ·never mined never read tengu_functional_llama ◇ Overrides the messages shown when Claude Code will continue automatically after hitting a usage limit. Code default null keeps the built-in text, such as Usage limit reached · continuing automatically {when} · esc to cancel. The value is an object with optional string fields waiting, started, rearmed and credits; an override for the first three is used only if it contains {when}, which is filled with at <time>, or with shortly (started, rearmed) or when it resets (waiting) when no time is known. Every message is capped at 120 characters.v2.1.284 → v2.1.306 ○gone never read tengu_gable_onyx_sluice ◇ Publishing a workshop page would run a structural check and refuse instead of shipping violations.v2.1.219 → v2.1.308 ◆in the build AND v2.1.248 tengu_garnet_finch ◇ Default false. Turns on the loop-command-nudge tip advertising /loop, which has a 3-session cooldown, for users signed in with a claude.ai subscription (OAuth with the user:inference scope); for everyone else the tip is eligible whatever the flag says. Either way the tip is hidden in a remote session (CLAUDE_CODE_REMOTE or a remote workspace) and needs cron enabled (CLAUDE_CODE_DISABLE_CRON unset and the tengu_kairos_cron flag on).v2.1.141 → v2.1.179 ○gone AND v2.1.308 tengu_garnet_loom ◇ No description from anybody.never mined ·never mined never read tengu_garnet_plover ◇ No description from anybody.never mined ·never mined never read tengu_gault_kestrel ◇ Default false. When on, the system prompt's guidance on hard-to-reverse actions ends its "Before deleting or overwriting, look at the target" sentence there, dropping the follow-on "If what you find contradicts how it was described, or you didn't create it, surface that instead of proceeding". The shorter wording is also used when CLAUDE_CODE_GAULT_KESTREL is set, when the Opus 5 prompt bundle applies (unless tengu_fennel_godwit is on), or when the client data key of this name is true.v2.1.257 → v2.1.259 ○gone AND v2.1.215 tengu_gb_eval_authed_enable ◇ Feature-flag requests can now attach auth asynchronously, but that path stays off in this build.v2.1.212 → v2.1.226 ○gone AND v2.1.227 tengu_gb_refresh_interval_minutes ◇ Feature flags now refresh in the background every six hours instead of only at startup.v2.1.221 → v2.1.308 ◆in the build AND v2.1.221 tengu_gentle_brooks ◇ Default true. When remembered MCP discover verdicts are on (tengu_dreamy_sundae), a value of false stops Claude Code reading or recording a verdict for HTTP toolbox MCP servers, whose cache keys start with ccr: and are built from the toolbox_mcp_server_id and mcp_url query parameters. A verdict records that the server spoke the current modern protocol version and is honoured for 24 hours; claude.ai proxy servers are unaffected.v2.1.286 → v2.1.308 ◆in the build never read tengu_gentle_frost ◇ A new SuggestPluginInstall tool renders an inline card offering to install a plugin found by searchv2.1.280 → v2.1.303 ○gone AND v2.1.280 tengu_gentle_grove ◇ Claude Code adds two new switches the server can turn on or off, one of them on unless switched off remotelynever mined ·never mined AND v2.1.304 tengu_gentle_hammock ◇ Remote Control requests can be gzip-compressed, but it is switched off in this build.v2.1.247 → v2.1.308 ◆in the build AND v2.1.308 tengu_gentle_hummingbird ◇ The control protocol gains a folder-listing request and a set of requests that let desktop, mobile and VS Code apps display and operate Claude Codev2.1.285 → v2.1.308 ◆in the build AND v2.1.293 tengu_giggly_dragonfly ◇ With input closed and no budget cap, a session with armed monitors can be held open past the background-task ceiling.v2.1.257 → v2.1.308 ◆in the build AND v2.1.295 tengu_git_created_prs ◇ New telemetry event tengu_git_created_prs counts pull requests printed versus announced during git operationsnever mined ·never mined never read tengu_git_operation ◇ Claude now re-checks later whether your pull request was actually created, instead of giving up after one try.never mined ·never mined never read tengu_glacier_2xr ◇ No description from anybody.never mined ·never mined never read tengu_gleaming_creek ◇ For MCP servers whose host is flagged by the cloud host check, Claude Code now trims the features it says it supports when connectingv2.1.305 → v2.1.308 ◆in the build AND v2.1.305 tengu_gleaming_cupcake ◇ New gates let Claude Code report tools that exist but are switched off as "available if turned on"v2.1.275 → v2.1.308 ◆in the build AND v2.1.275 tengu_gleaming_fair ◇ Resuming a session can now reuse an already-computed summary instead of recalculating itv2.1.138 or earlier → v2.1.308 ◆in the build AND v2.1.265 tengu_gleaming_fair_reuse ◇ Resuming a session can now reuse an already-computed summary instead of recalculating itv2.1.265 → v2.1.308 ◆in the build AND v2.1.265 tengu_gleaming_rivest ◇ A program asking Claude Code to re-initialize a session can now turn on prompt suggestions, if a remote setting allows itv2.1.285 → v2.1.287 ○gone AND v2.1.288 tengu_glimmering_glade ◇ Remote control re-registers and reattaches your sessions if the server forgets this machine.v2.1.242 → v2.1.308 ◆in the build AND v2.1.308 tengu_glittery_fern ◇ The skill-proposal tool can now refuse a proposed SKILL.md that keeps too few lines of your current one, using a threshold set remotelyv2.1.290 → v2.1.308 ◆in the build AND v2.1.290 tengu_glittery_mitten ◇ Stop hooks can now act on a kind of end-of-turn request that comes from an MCP server, behind a server-controlled switchv2.1.302 → v2.1.308 ◆in the build AND v2.1.302 tengu_glob_with_rg ◇ No description from anybody.never mined ·never mined never read tengu_glowing_lark ◇ Claude Code's rules for background tasks gained a launch-in-background-by-default value, next to a remote setting that is off by defaultv2.1.299 → v2.1.308 ◆in the build AND v2.1.299 tengu_glowing_orbit ◇ Inline shell commands in a skill's instructions can now be handed to the model to run explicitly instead of executing silentlyv2.1.271 → v2.1.308 ◆in the build AND v2.1.271 tengu_goal_checkin_injected ◇ Goals now nudge Claude every 30 minutes when background shells or agents are blocking evaluation.never mined ·never mined never read tengu_goal_cleared ◇ Replacing a live goal now clears the old one and records why it ended.never mined ·never mined never read tengu_goal_interruption ◇ Autonomous 'goal' turns get structured pause/retry/stop handling with clear 'Goal paused' messagesnever mined ·never mined never read tengu_goal_proposal_available ◇ Claude Code now logs once per session whether its goal-proposal tool was available to it.never mined ·never mined never read tengu_goal_proposal_decided ◇ Claude could propose a session goal you approve with one keypress, but it is switched off.never mined ·never mined never read tengu_goal_proposed ◇ Claude could propose a session goal you approve with one keypress, but it is switched off.never mined ·never mined never read tengu_goal_restored_on_resume ◇ Restoring a goal on resume records more detail, including when it replaced an existing goal.never mined ·never mined never read tengu_golden_popcorn ◇ CLAUDE_CODE_GZIP_REQUEST_BODY_LEVEL sets a gzip level from 1 to 9 for requests Claude Code already compressesv2.1.282 → v2.1.299 ○gone AND v2.1.282 tengu_goofy_hare ◇ A switched-off change lets Claude Code recover kept messages from an earlier compaction marker when the usual one is missingv2.1.301 → v2.1.308 ◆in the build AND v2.1.308 tengu_goofy_jellyfish ◇ In Claude Desktop, auto-compact on 1M-context models can now be capped at a 600k or 500k window by a server-side settingv2.1.283 → v2.1.306 ○gone AND v2.1.283 tengu_gorse_fathom ◇ Default false. When on, the auto-memory instructions gain a ## Project skill upkeep section: after saving a feedback memory that corrects a repeatable step, Claude is told to fold the correction into the matching .claude/skills/<name>/SKILL.md. It may only edit existing skills, except verify, which it may create.v2.1.198 → v2.1.308 ◆in the build AND v2.1.308 tengu_gorse_plover ◇ New internal flags added, including bashActFirstEnabled and three codename gates amber_astrolabe, bison_cairn, larch_cisternv2.1.248 → v2.1.287 ○gone AND v2.1.275 tengu_gorse_pylon ◇ Claude can post a quick 'on it' note in artifact comment threads before the real reply, off unless enabled.v2.1.232 → v2.1.308 ◆in the build AND v2.1.237 tengu_gorse_sill ◇ A second fast-ack mode lets Claude pick a fixed numbered acknowledgment instead of writing one, off behind two gatesv2.1.235 → v2.1.308 ◆in the build AND v2.1.235 tengu_gouda_loop ◇ Default false. When true, Claude Code checks for your closed issues on anthropics/claude-code: it reads a cached list and, at most once per 24 hours, runs gh issue list -R anthropics/claude-code --author @me --state closed, then shows a low-priority notice such as Your issue #N has been closed. Thanks for reporting! for issues not yet acknowledged and records them as acknowledged.v2.1.138 or earlier → v2.1.245 ○gone AND v2.1.308 tengu_gravel_chorus ◇ A new default-on internal flag was added with an unclear purpose.v2.1.260 → v2.1.308 ◆in the build AND v2.1.260 tengu_greedy_crane ◇ Claude Code checks a new remotely controlled switch and reads its settings from a renamed one, though what either controls is unknownv2.1.287 → v2.1.308 ◆in the build AND v2.1.287 tengu_grey_step2 ◇ No description from anybody.never mined ·never mined AND v2.1.68 tengu_grey_wool ◇ No description from anybody.never mined ·never mined never read