You can now use the Fable 5.2 and Mythos 5.2 models in Claude Code with the short names fable-5-2 and mythos-5-2. Fable 5.2 is now listed as the latest model, and fable52 works as an alias too. Plugins can use a new tool.check hook to make tool permissions stricter, but never looser for built-in tools. Plugins can also answer or rewrite outgoing web requests, and they are never lent your sign-in. Setting CLAUDE_CODE_ULTRAREVIEW_PR_AS_MERGED asks /ultrareview to review a pull request as it was merged. This release has 418 entries, and 270 of them are usable today.
Some features are in the build but are not switched on yet. A remote switch can make /privacy-settings open your privacy settings on claude.ai in your browser. Video attachments can carry a preview image, but only in hosted remote environments with a switch on. A skill's compatibility note can tell Claude when the skill needs your own computer or network, behind a switch that is off by default. Code editors may show a recommended extension, also behind a remote switch.
Stalled secure connections now stop with a clear error instead of hanging. The usual limit is 120 seconds, and the error says to check your network or proxy. The bridge command now loads the settings file you pass with --settings. Plugin sync now keeps plugins you already installed when your account's plugin list fails to load. A skills folder created after startup is now picked up and watched. The email address privacy prompt for shell and web actions no longer appears in the list of features.
Written by our agent from the shipped bundle, not by Anthropic.
Claude Code recognises the Fable 5.2 and Mythos 5.2 models, needs an explicit opt-in for fast mode on them, and keeps your model when fast mode supports it
Sessions & agentsclaude remote-control --settings <file> passes a settings file to spawned sessions, which refuse to start if that file or their start folder changed
A fable52 model alias is added, and the model picker gains a collapsible 'More models' row with Fable 5.2 highlighted
Claude Code can now check and merge a list of blocked MCP tools, called mcpDeniedTools, into its settings
ExtensionsPlugins can now hook outgoing web requests to answer or rewrite them, but are never lent your sign-in
ExtensionsWant the reasoning? Read walks the 54 entries that probably matter to you, each one opening to what changed and why.
Read this release → Every row →Anything you can use today, anything that visibly changes, and anything worth poking at. One line each, open for detail.
Claude Code recognises the Fable 5.2 and Mythos 5.2 models, needs an explicit opt-in for fast mode on them, and keeps your model when fast mode supports it
Unclear Whether these models are available to use is not stated.
claude remote-control --settings <file> passes a settings file to spawned sessions, which refuse to start if that file or their start folder changed
A fable52 model alias is added, and the model picker gains a collapsible 'More models' row with Fable 5.2 highlighted
Unclear It is not clear whether fable52 can be typed anywhere you choose a model, such as the model setting.
A mod blocking a request now gives a specific hook_denied error, notes you did not write are flagged as untrusted, and a few messages change
Unclear How mods are turned on, and whether an http.fetch hook can be set up in this build, is not stated.
Claude's EndConversation tool gets new guidance wording, a two-step confirmation before ending, and a recorded cause when it ends the session
Behind a switch, /privacy-settings can send you to claude.ai, and apps embedding Claude Code can record a privacy checkup as shown or dismissed
Claude Code's hook handling gains groundwork for hook modules: module-based plugins can count as holding hooks, and module hooks depend on a module capability
Unclear It is not stated what the extra step does to your hooks.
Claude Code can now check and merge a list of blocked MCP tools, called mcpDeniedTools, into its settings
Unclear It is not clear where the mcpDeniedTools list comes from or whether anything uses it yet.
Plugins can now hook outgoing web requests to answer or rewrite them, but are never lent your sign-in
Unclear It is not known whether plugins can use this hook by default.
A new tool.check hook lets plugins answer allow, deny or ask on a tool check, and only stricter answers take effect for built-in tools
Unclear Whether the plugin system that runs this hook is switched on by default is not known.
Plugin ui.notify gains delay, time, until-back and key options, a new ui.cancelNotify removes notices, and headless sessions pass them to the host
Unclear Whether Claude Code acts on a ui.cancelNotify message in this release is not shown.
Programs driving Claude Code can block specific MCP tools per session, and network errors now separate unreachable hosts from policy blocks
CLAUDE_CODE_YOU_SHOULD_KNOW_AUTO_SEND environment variable#Claude Code now reads a CLAUDE_CODE_YOU_SHOULD_KNOW_AUTO_SEND environment variable
Unclear What setting CLAUDE_CODE_YOU_SHOULD_KNOW_AUTO_SEND actually turns on, and its default, are not settled.
Setting CLAUDE_CODE_USER_DELEGATED_SESSION_DIRECT_API gives remote sessions a separate failure type when their route is refused
Two user delegated session variables are passed to spawned processes, and --bg and plugin eval refuse to run when the direct API one is set
Unclear It is not clear what values CLAUDE_CODE_USER_DELEGATED_SESSION_ALLOWED_CHECK_MODELS accepts or which models it lets through.
Claude Code now has messages suggesting Defense Access or the Cyber Verification Program to reduce interruptions to security work
Setting CLAUDE_CODE_GIT_CHECK_ONLY_IN_REPOSITORY lets Claude Code skip its built-in git check when you are not inside a git repository
Unclear It is not confirmed that setting the variable actually stops the hook from running.
Setting CLAUDE_CODE_CCR_REQUESTS_BY_FETCH on a hosted worker sends unproxied network requests through Bun's built-in fetch
MCP servers named in CLAUDE_CODE_TOOL_PICTURE_SERVERS can mark pictures as shown to you, and Claude is told it need not send them
Unclear How server names should be separated in CLAUDE_CODE_TOOL_PICTURE_SERVERS is not stated.
Programs driving Claude Code through the SDK can now set autoModeVerdictTools when registering an MCP server
Unclear What value autoModeVerdictTools expects is not stated.
Setting CLAUDE_CODE_ULTRAREVIEW_PR_AS_MERGED asks /ultrareview to review a pull request as merged, or refuses if that is unavailable
Unclear Whether reviewing a pull request as merged is actually available to any account is decided by the server and is not known.
Auto mode now refuses an action with a clear message when none of the models allowed to check it can be used
When CLAUDE_CODE_USER_DELEGATED_SESSION_DIRECT_API is set, auto mode switches off a third-party checker role and always checks allowed MCP write tools
Unclear What sets this variable, and what the fallback does when the classifier refuses, is not shown.
Claude Code can fetch a privacy checkup notice from Anthropic, show it in the terminal or VS Code, and report whether you saw or dismissed it
Unclear It is not clear whether a remote switch decides when this notice is shown.
The description of transcript_allowed now names Pro, Max and Ultra as the individual plans it applies to, where it named only Pro and Max
Unclear Only the description wording changed, so it is not known whether Claude Code now treats Ultra accounts differently.
Claude Code can now tell a code editor to show an extension recommendation, only when a remote switch is set to offer it
Unclear It is not clear what extension or product the offer recommends.
Claude Code now contains message formats for offering a recommended extension, including a check on idle time before showing it
Unclear It is not known whether any screen shows this offer yet, or what turns it on.
A skill's compatibility note can be shown to Claude, so it says plainly when a skill needs your own computer or network that it cannot reach
Claude Code can now ask you which computer to use before linking one to a chat, and denies the request if claude.ai does not answer in time
Claude Code's Vertex AI model check has a new bearer-token sign-in mode, but nothing in this build can switch it on
Uploading a video attachment can now include a thumbnail image, only in hosted remote environments and only when a remote switch is on
The no_reply_needed tool now belongs to a group of tool names that also includes the research report and plan tip tools
Unclear What this group of tools is used for, and what makes the tool available, is not stated.
A stop, cancel or Esc can now end the current turn without stopping running background agents and workflows
theme.set, not only the backdrop#The plugin theme.set call now accepts colours as well as a backdrop, and needs at least one of the two
CLAUDE_CODE_MCP_SERVE_GARBAGE_COLLECTIONS controls when mcp serve frees unused memory, and slow cleanups switch it off
Unclear It is unclear whether the default cleanup schedule applies when the variable is not set.
The apps gateway now refuses to serve a desktop policy entry with unknown fields that Claude Desktop would drop or misread, instead of only warning
When the Chrome extension is not connected for a request from mobile, the message now offers a cloud browser as well as web search
In a user-delegated session, Claude Code now refuses API calls to routes that are not allowed and shows an error
Unclear It is not clear what makes a session user-delegated, so it is unclear who is affected.
Text wrapping a voice-call record now tells Claude it is unconfirmed background, may be old, and is neither a request nor approval
Claude Code has two new remotely controlled checks, off by default, one adding a stricter check on plugin entries served over MCP
Unclear What tengu_hazy_boole controls is not shown.
Claude Code's pricing table now has a tier at $8 input and $40 output per million tokens, with matching cache prices
Claude Code now has saved settings and event names for an organisation model-choice move, goal re-arming and plugin restore, but none do anything visible yet
The plugin hook system lists a new "ask" decision, new guard categories, and refuses hook paths that are not absolute
Unclear Whether the new "ask" decision and guards already take effect, or exist only as names so far, is not settled.
When a plugin hook rewrites a tool's description, Claude Code now also keeps the hook's cache-reset marker and whether the tool loads later
claude plugin restore-set-aside <plugin> gives set-aside plugin options and secrets to the current plugin, or deletes them with --discard
Plugin notifications now take a key, a delay or time, and an until-back option, and can be cancelled
Unclear It is not clear whether plugins of this kind can be used by everyone in this release.
Updating MCP servers from a program can now include tools to deny, and servers waiting for sign-in are retried afterwards
Unclear The retry of servers waiting for sign-in sits behind a check whose setting is not known, so it may not run for every session.
A new autoModeVerdictTools setting lets SDK-supplied MCP tools be marked as already allowed when the server-side auto mode check approved the exact call
Unclear It is not shown what Claude Code does with a call once it is marked as allowed by the server.
Setting CLAUDE_CODE_WEBSEARCH_CITATIONS is now enough to get citations in web search results, with no remote switch also required
Remote Control now accepts sessions whose folder was chosen on claude.ai, checking the folder first and refusing with a message when it cannot
Unclear What controls whether subfolder support is turned on was not identified, so it is unclear who can use it.
claude.ai can ask Claude Code to list folders on your computer to start a session, following your Read deny and ask rules
Unclear It is not clear what switch turns this on or whether it is on by default.
claude remote-control now takes --settings with the path of a JSON settings file, and rejects inline JSON text
--settings#Starting the bridge command with --settings now loads that settings file, which the command used to ignore
Claude Code's message for a refused connector file download now says the limit is 500 MB, up from 100 MB
Unclear Only the message wording is known to have changed, not that the actual size limit was raised to match.
What would fix it?
Smaller changes and internals, grouped as the pipeline found them. Nothing is dropped, it is only further down.
18 more of these are in What probably matters to you, above.
project_write refuses writes whose if_version does not match what the session saw, gains new failure codes, and reads local files with a configurable limit
Tool calls a plugin makes now skip auto mode's classifier when plugin policy allows them, and are refused when they would need a person's approval
Unclear Whether plugin tool calls now end up allowed or asked more often than before is not shown.
Plugin MCP servers matching a connector your organization blocked on claude.ai can now be blocked or just counted
Unclear It is not settled which mode applies when nothing sets require_connector_block_on_plugin_servers, so whether blocking is enforced by default is unclear.
Uploaded attachments can now include a JPEG thumbnail preview, and the stopped-task reminder simply says the task was stopped
Bedrock, Mantle, Foundry and Vertex now look up a credential first and use it as a bearer token or API key before the usual sign-in
Unclear Where the credential lookup gets its values from, and so how you would supply one, is not clear.
The gateway no longer reads the Bedrock, Vertex or Foundry base-URL variables, and Bedrock without base_url uses its regional endpoint
The upgrade option shown when you reach a usage limit now reads "Upgrade to Max or Ultra" instead of "Upgrade to Max"
When an option is set, compaction can make a second request summarizing its summary with the kept messages, keeping the first if that fails
Unclear It is not known when the option that turns on the second summary is set.
Claude can get a new privacy section and stricter wording about keeping your email private, and the built-in email check now blocks instead of asking
Unclear What decides which version of the email wording a given session gets is not known.
Claude Code can warn you when your version is below an announced minimum, and suggests asking your admin or running claude update
Claude Code reports when mods are reloaded, held, kept out, unbound or failed, and in bypass or auto mode, mods Claude writes turn on without asking
Before building slides or a design, Claude is told to ask three or four tailored questions, with a look picker and a design-system option
Unclear What switches the question step on in this release is not settled.
fable alias at startup#If your saved model is Fable 5.1, Claude Code switches it to the fable alias, with a delayed move for team and enterprise organisations
A built-in privacy hook, on by default, refuses shell commands that contain your account email and also set a User-Agent
Unclear What decides whether this module is available on a given setup is not stated.
Claude is told fast mode may run on your current model or switch to Claude Opus, and the 'Fast mode OFF' status now points to /fast
If your organization sets an upcoming minimum version and date, Claude Code can show a notice before that version becomes required
The main system prompt Claude Code sends when it compacts a long conversation was rewritten
Unclear The new wording of the prompt is not shown, so what it changes in compaction is unknown.
When the server has set your Project, /project now says it cannot change it and that it lasts until the session ends
On macOS, Claude Code deletes the ClaudeCode.app helper at startup when it does not use it, unless a process is still running from it
In remote sessions, a model alias can now resolve to the newest model in the same family
A new warning tells mod authors that a hook which never passes control on hides what other mods draw
In sessions started from an artifact launcher, Claude can now carry on when the host reports nobody answered its question
In remote sessions, certain slash commands now reply "This command isn't available in a chat." instead of running
Unclear Which commands are affected and what makes a session count as a chat are not known.
A notice tells Claude that a background agent is being resumed automatically from a saved transcript that may be missing its last steps
Claude Code only accepts a keep-percent value in frontmatter if it is a whole number between 50 and 100
Unclear Which file or feature reads this keep-percent frontmatter value is not stated.
Some plugins with no saved on or off choice can now count as enabled when a further condition is met
Unclear It is not clear which plugins this affects or what the two extra checks test.
Claude Code now warns when a subagent's output was not reviewed, with wording that depends on CLAUDE_CODE_USER_DELEGATED_SESSION_DIRECT_API
Unclear What else setting CLAUDE_CODE_USER_DELEGATED_SESSION_DIRECT_API changes beyond this warning is not known.
Stopping a local workflow now tells Claude it was paused and that you can resume it from /workflows
When neither the environment variable nor the model's own setting gives a seconds value, the silent-turn reminder now takes a default from a remote check
Unclear The default number of seconds the remote check supplies is not known, so it is unclear whether the reminder now fires for people who set nothing.
When Claude Code runs without an interactive terminal, a new startup check can refuse a user-delegated session and exit
Unclear What the startup check tests, and so when a start is refused, is not stated.
With CLAUDE_CODE_HOST_OWNED_CONFIG_DIR set, a ~/.claude.json that cannot be read no longer stops Claude Code; it starts on defaults
If a memory file Claude read is gone from the memory listing, Claude Code can add a notice that the earlier read is stale
Unclear It is not clear which sessions this stale-read check applies to.
A request from a Remote Control client to change a session's working directory can be refused with the reason remote_origin
MCP tool calls can send download progress, tools can be marked as delivering files, and the UI handles download updates
Unclear What Claude Code does differently with a tool marked as delivering files is not shown.
Claude Code's startup report to SDK hosts now flags sign-in conflicts and hook-ask priority, and the upload approval prompt mentions .xlsx files
Unclear It is not clear whether .xlsx files are now covered by an earlier approval or still ask before every upload.
MCP tools now record where their server came from, and alwaysLoad only counts for directory or custom servers when further checks pass
Unclear It is not stated where Claude Code applies this rule, so its effect on servers that set alwaysLoad today is unclear.
Claude Code's shell setup now wraps pkill, pgrep and pidwait so they add --ignore-ancestors when supported, keeping them from matching the shell running them
Unclear Which platforms get these wrappers is not stated.
After a session restarts, its artifact comment monitors can be turned back on when the session's own job vouches for them; others raise a consent notice
Changing effort with the keyboard now shows a dim hint in the prompt box when the level in effect differs from the session's
Claude Code's built-in email check now refuses a tool call carrying your email address rather than asking, and also spots the encoded form %40
When a session runs in dontAsk permission mode, Claude Code refuses to load plugin mods and hooks
Claude Code skips connecting to an MCP server when the headers it would send fall under an organization's connector block
Unclear Where the organization's connector block is set, and which users it applies to, is not stated.
When a conversation is compacted, critical reminders are kept and very long reports from other agents are shortened or saved to a file
Compaction now passes over voice-call messages when finding your last turn, and the Chrome permission prompt reports when a tab's address can't be found
A plugin's MCP server sign-in is stopped when Claude Code cannot check it against your organization's connector settings
Claude Code now skips synced plugin MCP servers that offer no way to register for sign-in, without needing a remote switch first
Published verbatim by Anthropic for v2.1.321. Text is unmodified from the upstream changelog. Everything else on this page came out of the bundle instead, which is why the two lists don't match.
Of these 118 bullets, 23 name something an entry on this page also names, 30 name something no entry here does, and 65 name nothing specific enough to line up either way. The pairings are made on names both sides wrote down, a flag or a setting or a slash command, so read one as probably the same thing rather than as a fact, and read the middle number as candidates rather than as a miss count.
claude-fable-5-2), now the default Fable model — 1M context, $8/$40 per Mtok with $0.20/Mtok cache reads
Probably fast-mode-now-needs-an-explicit-opt-in-on-fable-52-mythos, fable-52-and-mythos-52-model-recognition-added --settings <file> to claude remote-control, which applies a settings file to every session it starts; if the file changes, no new session starts until you restart claude remote-control
Probably cloud-sessions-refuse-an-automode-block-instead-of-dropping, remote-control-refuses-settings-before-the-verb-with-a-hi, remote-control-settings-forwarded-to-child-sessions-dire, remote-control-child-sessions-refuse-a-settings-file-that-ch, bridge-fast-path-now-honours-settings, remote-control-cli-adds-settings-and-debug-file-rejects, remote-control-gets-a-settings-flag-for-spawned-sessions cp or mv into a directory writes
Nothing to match on claude mcp login printing a SEP-2352 issuer warning each time you sign in again to a server that Claude Code registered with
No entry names this #108158[BUG] MCP OAuth: SEP-2352 "no 'issuer' stamp" warning on every claude mcp login — provider doesn't round-trip issuer Open
NotebookEdit deleting or overwriting a notebook's first cell when the call named no cell, on network paths (UNC shares, WSL folders, /net mounts)
No entry names this old_string, or editing a notebook as text, on network paths (UNC shares, WSL folders, /net mounts)
No entry names this /config showing Manual as the default permission mode when new sessions start in auto mode; pressing Space on "Auto (default)" now saves Don't Ask, not Plan
No entry names this FileChanged hooks stopping for paths set through watchPaths after the first shell command, when the session's working directory had a lowercase drive letter
No entry names this --plugin-dir . and --plugin-dir ./builtin writing generated type files for Claude Code's built-in plugins into a builtin folder
No entry names this /compact showing twice in Remote Control on web and mobile
No entry names this #100427[BUG] Remote Control: one /compact shows twice in the phone app (typed prompt and command-name record both rendered) Open
/plugin marketplace updates downloading a plugin again for each worktree or folder of a repository
No entry names this git stash, including after entering a worktree
No entry names this claude -p sometimes telling Claude that background commands get 30 minutes when they are stopped after 10
Probably headless-p-error-exit-carries-a-reason #103473[BUG] Print mode: Bash schema tells the model background commands get 30 min, but they are killed at 10 min (and the stated default flips between launches, breaking the prompt cache) Open
pkill, pgrep and pidwait in cloud sessions matching the session's own parent processes
Nothing to match on ask reason instead of another hook's when both asked
Nothing to match on ask being lost when another hook's defer finished first and could not be honored
Nothing to match on /desktop on macOS saying Claude Desktop is not installed when it is in ~/Applications
No entry names this #107903/desktop does not detect Claude Desktop installed in ~/Applications (still reproduces on 2.1.319, same as #41899) Open
#98476[Bug] Desktop command fails when claude.app installed in ~/Applications instead of /Applications Open
vcs.*) missing from Claude Desktop sessions that start on a pre-started process
No entry names this #108072[BUG] Desktop app sessions never report vcs.repository.* OTel attributes (spare process starts outside the repo) Open
claude agents ignoring --permission-mode, --dangerously-skip-permissions, --model, --effort or --agent typed before agents when a flag such as --verbose was also on the line
Probably remote-control-refuses-settings-before-the-verb-with-a-hi, remote-control-gets-a-settings-flag-for-spawned-sessions claude -p, SDK and VS Code sessions hanging, when memory ran out while the transcript was being read
Probably headless-p-error-exit-carries-a-reason grep -r foo "$PWD/src" asking for permission when the same command with the path typed out runs without asking
No entry names this desktop setting holds a field newer than the gateway, such as a managed MCP server's oauth.omitResource; it now warns and serves it
No entry names this .zshrc sets KSH_OPTION_PRINT
No entry names this #96333[BUG] Bash tool produces no output when KSH_OPTION_PRINT is set in ~/.zshrc Open
UserPromptSubmit hook blocked a replayed permission answer; these hooks no longer run on that replay
No entry names this claude remote-control accepting a y typed without Enter at its first-run question when the terminal's input ended
Probably remote-control-refuses-settings-before-the-verb-with-a-hi, remote-control-settings-forwarded-to-child-sessions-dire, remote-control-cli-adds-settings-and-debug-file-rejects code: rules
Nothing to match on PreToolUse mcp_tool hooks on stdio and HTTP MCP servers being skipped when their answer was over 16 MiB; the tool call is now refused
Probably remote-mcp-hook-calls-get-clearer-errors-on-oversize-results claude -p on Linux exiting 0 in silence when a full NFS export or a used-up NFS quota lost transcript lines: it now prints an error and, in text and json output, exits 1
Probably headless-p-error-exit-carries-a-reason claude mcp list output, when a stored MCP sign-in record was null
No entry names this /model picker: "More models" is now an unnumbered row below every other, and older models open under it without renumbering the rest
Probably remote-control-slash-commands-unavailable-here-get-specific, model-switch-notice-while-the-cloud-session-is-not-ready, model-picker-gets-a-collapsible-more-models-row-and-fable /device page: they no longer name gateway settings, which the gateway's own log still names
No entry names this claude install and the install scripts: they finish 2 to 4 seconds sooner
No entry names this claude --bg --worktree session that couldn't set up its worktree: it now shows the error from its beginning, not an internal tag or its last 200 characters
Probably worktree-startup-errors-routed-through-a-common-exit-helper /login is offered only if it's one of yours, and Claude now says where a copy may be after an account move
Probably remote-control-other-org-message-reworded, artifact-error-for-other-org-now-suggests-login, artifact-watch-error-messages-reworded, artifact-lookup-messages-cover-organizations-the-user-has-le claude remote-control's messages for an unanswered first-run question: each now says what to do next
Probably remote-control-refuses-settings-before-the-verb-with-a-hi, remote-control-settings-forwarded-to-child-sessions-dire, remote-control-cli-adds-settings-and-debug-file-rejects allowedProviders setting description to name Ultra alongside Pro and Max, and the forceLoginMethod description to cover every paid Claude plan
Probably managed-settings-read-failures-now-re-read-and-report-reason, login-method-setting-description-reworded allowUnsandboxedCommands: false or allowManagedDomainsOnly, and not to load in dontAsk mode
Probably hooksmods-refused-in-dontask-permission-mode $.tool.call, $.mcp.call) to raise no permission prompt, even under your own ask rules or in dontAsk and -p runs; deny rules and organization policy still apply
Probably hooksmods-refused-in-dontask-permission-mode fable: on Pro and Max plans at once, on Team and Enterprise plans once your organization allows it; use /model claude-fable-5-1 to switch back
Probably remote-control-slash-commands-unavailable-here-get-specific, model-switch-notice-while-the-cloud-session-is-not-ready, model-picker-gets-a-collapsible-more-models-row-and-fable /fast, /config and /model to name your current model when it supports fast mode, and removed "(Opus only)" from the VS Code row
Probably remote-control-slash-commands-unavailable-here-get-specific, model-switch-notice-while-the-cloud-session-is-not-ready, model-picker-gets-a-collapsible-more-models-row-and-fable, fast-mode-prompt-text-changed-no-longer-says-it-never-downg, fast-mode-off-status-line-hints-fast .claude/skills folders inside ~/.claude/plugins no longer load when Claude works on files there
No entry names this num_hooks and num_success after Bash and PowerShell calls
No entry names this AskUserQuestion in claude -p and SDK sessions to deny a host's permission answer that sets the undocumented afkTimeoutMs input field
Probably headless-p-error-exit-carries-a-reason, askuserquestion-timeout-can-be-admitted-as-nobody-answered, server-supplied-askuserquestion-prompt-text-removed-tengu-c extraKnownMarketplaces: a declared name can't be added from another source, an entry from elsewhere is hidden, and plugins the declared marketplace lists are reinstalled from it
No entry names this policy_mode: layered in the Claude apps gateway: managed MCP servers now add up over every matching policy, match: {} included, so a gateway already on it serves more after the upgrade
No entry names this ANTHROPIC_*_BASE_URL variables in the gateway's environment. Set base_url on the upstream to use another endpoint
Probably gateway-warning-for-base-url-env-vars-changed-and-now-covers, bedrock-requests-always-get-an-explicit-regional-base-url .claude/skills when their files change; set CLAUDE_CODE_PLUGIN_DIR_WATCH=0 to turn this off
No entry names this cp and mv spelled by a path, such as /bin/cp, to get the same path checks as cp and mv
Probably telemetry-and-config-additions extraKnownMarketplaces take-overs from an undeclared source: saved data is set aside for cleanupPeriodDays, and options and secrets until claude plugin restore-set-aside
Probably new-claude-plugin-restore-set-aside-command-with-disca A model matched these bullets to the GitHub issues they fix, so a link can be wrong.
2 added and 2 removed, of 218 lines, about 38 words, in the prompt 14 of 27 arms receive. 4 other prompts also changed. The appended system-reminder blocks moved: 2 lines added, 2 lines removed.
Claude Code, interactive mode
5 prompt changes in this release could not be quoted from the build, so no entry on this page describes them.
834 documentation changes were recorded within 24 hours either side of this release, nearest first. The closest 12 are below. They're here because they happened near this release in time. That's not a claim that this release caused the edit, or that the page documents anything in it.
The 57 literal strings found in the bundle, with the number of entries that name each one. Picking one searches for it. A name is here because this build's code mentions it, which is not the same as it working or being finished.
What's wrong with this entry?