You can now use Sonnet 5.5 by choosing sonnet on the Anthropic API, where it replaces Sonnet 5. Bedrock, Vertex and other providers still get their older models under that name. A new sandbox network setting, deniedResolvedAddresses, blocks allowed websites that resolve to IP addresses or ranges you list. Apps built on the SDK can save a feedback report as a local zip with the new save_locally option. The Claude gateway can now sign in to its identity provider with a certificate, so oidc.client_secret is optional. /rate-limit-options now shows up in the command list.
Several features are in this build but not switched on yet. Claude Code can clear old tool results when you return after more than an hour idle. It keeps the last 5 results and stays off unless switched on remotely. PR Steward is an agent that looks after pull requests. /autofix-pr can now step aside when PR Steward is already working on one. A hidden guide plugin for the live mods added in v2.1.283 is built in but off.
Cleaning up an agent team no longer deletes its members' worktrees. Ending a session now closes only the tmux or iTerm teammate panes that session opened. Calls to an MCP tool whose server is still starting now wait instead of failing. A broken image in a prompt no longer crashes the turn. The /design canvas mode and the dashboard artifact template are removed. Claude can also no longer start /design on its own.
Written by our agent from the shipped bundle, not by Anthropic.
The sonnet alias now means Sonnet 5.5 (claude-sonnet-5-5) on the Anthropic API, Sonnet 5 stays pickable, and refusal handling now covers Sonnet
Sessions & agentsRemote sessions can move a hash-verified opening of --append-system-prompt text into the cached system prompt via CLAUDE_CODE_APPEND_PROMPT_HEAD
ElsewhereA built-in cc-plugin-otlp can send OpenTelemetry records through plugin hooks, and plugin telemetry hooks must name a stream, with anthropic reserved
ExtensionsA new hub row lets you see or withdraw consent for cloud sessions running commands on your computer, and policy-blocked Yes answers are not stored
Sessions & agentsSDK hosts can save a feedback report as a local redacted zip with save_locally, and learn at startup whether feedback is sent, bundled or disabled
Sessions & agents
--model help text no longer gives an example full model nameSessions & agents/design drops its design canvas mode, and Claude can no longer run it on its ownExtensions/design canvas option are goneTools & outputWant the reasoning? Read walks the 42 entries that probably matter to you, each one opening to what changed and why.
Read this release → Every row →9 more of these are in What probably matters to you, on page 1.
.claude/rules from outside the project now need external-include approval#A symlinked .claude/rules pointing outside the project is skipped unless external includes are approved, and the warning now names .claude/rules
Unclear The exact set of cases in which a symlinked rules folder is now skipped is not fully clear.
In -p and SDK runs, the first turn now waits for MCP servers named in allow rules or mcp_tool hooks before starting
Unclear It is not confirmed that the servers collected from allow rules and hooks are the same list the headless wait uses.
ANTHROPIC_FOUNDRY_RESOURCE must be a plain Foundry resource name#ANTHROPIC_FOUNDRY_RESOURCE is now checked as a 2-64 character resource name and rejected with a clear error if it is a URL or host name
Claude Code now refuses to read settings and other files through a symbolic link that leads onto a network share or automount path
Unclear The exact conditions that trigger the refusal are not fully known.
In a session that cannot write to disk, adding a claude.ai marketplace now fails with a clear message instead of attempting the save
Unclear It is not clear which kind of session counts as unable to write to disk, or what happened in that situation before.
When publishing an artifact to a type-locked instance fails, Claude Code now picks between two hint texts, or none, based on a setting
Unclear What the two hint texts actually say is not known.
If Claude calls a browser tool like read_page without its prefix, the error now names the correct prefixed tool to call
Unclear Which server prefixes count as known is not settled.
Attaching a git- or folder-backed synced session now needs a signing key for this device, and sync stops if there is none
Unclear Which users or sessions this sync feature is available to is not settled.
claude --cloud no longer lets an upload-only folder sync consent block git directory sync
Unclear Whether upload-only sync is now handled elsewhere or has been dropped is not settled.
The xhigh effort description now ends "on supported models" instead of naming a model, and the check for which models allow it was reworked
Unclear Whether the reworked check changes which models allow xhigh is not known.
Hooks from a plugin that runs on another machine now report that the machine is unavailable, instead of failing with an unexplained error
Unclear It is not clear what makes a plugin's hooks run on another machine.
In cloud sessions with no disk, prompt and agent hooks are told there is no transcript file, so they stop looking for one
Publishing to an Artifact made from an Artifact type is refused in this session, and Claude is told retrying won't help and to tell you
Unclear It is not clear which sessions have access to Artifacts at all.
When Claude Code is launched in diskless mode, your statusLine and fileSuggestion commands no longer run
Unclear How a host or user turns diskless mode on.
In sessions that run file and shell tools only in the cloud, Bash is never counted as read-only, and a local git check is skipped
Unclear It is not confirmed that the git status check and the hook change depend on the diskless option.
The debug log now shows a hook's normal output and error output under separate labels, plus its exit code when it fails
Tab headers in dialogs now wrap onto extra lines when the terminal is narrow, instead of staying on one row
Unclear It is not clear which dialogs let their content take over the arrow keys.
/auto-mode-setup checks you are signed in before it scans your environment#/auto-mode-setup now checks your login first and tells you to run /login if it has expired or you are not signed in
Unclear It is not clear whether the login check applies in every environment, since part of the condition that triggers it was not identified.
Running as root on Linux without the CAP_SETFCAP permission, the sandbox now reports a clear error instead of every command failing
Publishing an artifact from a file on a network share is now refused unless that location is listed as a trusted network directory
Unclear Exactly which network path forms are caught, and how trusted network directories are set, is not stated.
Under allowManagedPermissionRulesOnly, only plugins from admin-backed sources keep their allowed-tools grants; other plugins' grants are ignored
Unclear Only the setting's description changed visibly; the code that enforces it was not seen.
When Claude Code refuses to prune git worktrees, the message now says whether git failed, gave no answer, or the folder is owned by someone else
Gateway checks now warn when availableModels is empty or leaves out the default model, both of which get requests refused
Unclear Which command or screen shows these warnings is not confirmed.
Several tool calls in one response now share one wait for a connecting MCP server, and a server that already timed out can be skipped
Unclear When the skipping of already timed-out servers is switched on is not known.
A plugin that fails to load because of its manifest now shows whether the file could not be read or which checks it failed
Through a gateway provider, the spend limit line now shows money used against the limit, the period, the percentage and when it resets
Unclear It is not clear which command shows this line.
/loop runs confirm and re-arm in a different order#The instructions Claude follows when /loop paces itself were rebuilt, changing when it confirms, how it re-arms and what it says on stopping
Unclear It is not clear what decides whether the confirmation comes before or after the next run is scheduled.
When managed settings require a specific sign-in but another credential is set, the error names that credential, where it came from, and how to fix it
Artifact pages can now load scripts from unpkg.com, cdn.tailwindcss.com and code.jquery.com, as well as cdnjs and jsdelivr
With allowManagedPermissionRulesOnly set, npm-based marketplaces keep their plugins' allowed-tools only when a registry is pinned
Unclear The rule is stated in the settings descriptions, and where Claude Code enforces it is not settled.
When an MCP server asks you to sign in again for more access, a message now says the browser is opening and how to get the link
Missing, malformed or unreadable Google Cloud key files are now reported as Google Cloud credential problems when using Vertex
After scheduling a wakeup in /loop, Claude is now told to post any owed status update and then end its turn
Unclear What decides when the "only thing Claude does" wording is used is not settled.
Monitor events in background task notices now show as a one-line headline, and finished tasks say how many messages they sent
Claude Code now hides repeated turn-time notes when the same number of background agents and workflows are still pending
Unclear Which view shows every turn-time note is not known.
On /model switches, the effort level is now applied directly, including levels that came from ultracode, and ultracode state is no longer set
Unclear The overall effect a user will notice is not known.
The plugin runtime's $.process.run now reports isStdoutTruncated and isStderrTruncated when output passes its 4 MiB limit
Unclear It is not clear whether the 4 MiB limit is also new or only the two truncation values are.
The sandbox's HTTPS proxy now keeps connections open for reuse and includes a non-default port in the Host header it forwards
In fullscreen mode, hover highlights now update when content scrolls or shifts under a still mouse pointer
If GOOGLE_APPLICATION_CREDENTIALS points to a file that does not exist, Claude Code now names that variable and path in the error
With a configured model list, opus, sonnet and haiku now map to the first model of that family in the list, not the canonical claude- ID
Unclear It is not clear where the configured model list comes from or which features use this lookup.
Claude is now told to skip questions with an obvious default or an answer it can check itself, pick the obvious option, say so, and carry on
Unclear The prompt record lists the AskUserQuestion description as unchanged since v2.1.283, which conflicts with the added text reported between v2.1.283 and v2.1.284, so the release it arrived in is unclear.
Cloud-to-machine sync adds four withheld_* causes and a new reason saying your machine already has the work
Unclear It is unclear what this change on the warning path means for what you see.
--channels, allowedChannelPlugins and duplicate-server warnings now handle built-in marketplace plugins under their current and previous names
Unclear It is not clear which names the built-in name mapping connects.
Per-worktree git config now has an 'inert' state with its own error message, and the refuse policy now rejects any worktree config that is present
Unclear It is not clear which worktree configuration states exist besides absent, extension and file, or exactly when this check runs.
Directory sync now records each downloaded apply's written files and upload generation, and compares the shared .git directory's identity
Unclear It is not stated which visible Claude Code feature depends on this comparison.
When you refuse to let Claude save to account memory, Claude can now receive a message written for that dialog instead of a generic one
Unclear It is not clear whether account memory is available to users yet.
When a plugin's dependency has a version conflict, the error message can now suggest updating that dependency
Unclear The exact wording of the suggestion, and when it appears, are not known.
The Linux sandbox now mounts the parent folders of protected paths inside writable areas as read-only
Unclear The exact problem this guards against, such as renaming a parent folder, is not stated.
deniedDomains rules now also match a host written another way, such as a different IP address spelling or a trailing dot
Unclear Exactly which alternative spellings are recognised is not fully settled.
claude plugin eval ignores keys pressed too early at its trust question#The "Trust this plugin directory?" question from claude plugin eval now ignores answers typed too quickly and asks again
Unclear How long the window is was not determined.
More custom prompt commands now keep their full name as an alias, which used to happen only for names that contain a colon
Unclear It is not clear which names pass the new check.
Under allowManagedPermissionRulesOnly, plugin skills can keep their allowed tools based on official attestation or marketplace source
The artifact tool now explains when a link is a claude.ai chat artifact, chat or project link, or a bare id, and suggests what to do
Unclear It is not clear who can use the artifact tool in this release.
Gateway configuration now checks the auth settings on each telemetry.forward_to entry and reports problems with their Google login settings
Unclear Which specific problems the check reports is not stated.
The reader that checks shell commands before they run now follows backslash line breaks and marks more tricky commands as unreadable
Unclear What happens next with a command marked unreadable is not settled, though most likely Claude Code asks you for permission.
bin folders are added to PATH only if they exist#Claude Code now checks each plugin bin folder exists before adding it to PATH, and leaves out ones that are missing
Plugin options that are on/off switches now show true or false, using the default when unset, instead of showing blank
Unclear Which screen or output shows this list of values is not confirmed.
A plugin setting that should be true or false is now dropped if its value cannot be read as one, instead of being forced into one
Unclear Exactly which texts are accepted as true or false is not known.
On Linux, the sandbox glob warning now also checks filesystem.denyRead and filesystem.allowRead patterns that start at the root or have no fixed folder
/loop has most models schedule the next run before confirming#When /loop paces itself, most models are now told to decide and schedule the next run first and write their confirmation afterwards
Unclear It is not clear which models besides claude-mythos-5 keep the old order.
When a plugin install fails, Claude Code now undoes only what that failure left behind, and can suggest updating a dependency that is too old
Unclear How the suggestion to update a dependency is shown to the user is not settled.
The Bash tool's per-command allowed_domains check now also rejects disguised loopback or IPv4 addresses written inside square brackets
Unclear Exactly which bracketed addresses got past the check before is not settled.
A new message covers a cloud session that has none of this folder's files because this computer's device key could not be read
Unclear It is unclear where this message is shown or what decides when it appears.
The Artifact tool's instructions now accept only claude.ai artifact links and say a publish is refused unless Claude has read the artifact first
Unclear It is unclear whether the refusal is enforced in this version or only described in the instructions.
The 'Run this command now?' prompt when installing a plugin now discards any answer given within 250 ms and asks again
Unclear The same check is used at one other prompt, and it is unclear which one.
claude remote-control --help now describes the trust prompt and worktree needs#The claude remote-control help no longer says to run claude first to trust a folder, and notes what worktree mode needs
Sandbox deny paths that end in a slash now show a validation error, because they matched nothing and protected nothing
Unclear It is not clear whether this error stops the settings from loading or is only reported.
The notice that a plugin installs by running a command from a marketplace now goes to the confirmation prompt too, not only to the screen
Unclear It is not clear how the approval prompt displays the notice.
The prompt-suggestion back-off hint now only shows after a normal prompt, not after slash commands or words like exit and :q
Unclear One further condition that skips the hint could not be identified.
In the Save configuration form, a number field now warns about input that is not a number, and empty input is still accepted
Unclear It is not certain that this form is the one for plugin settings.
In the extra-usage buy flow, Claude Code now checks for a saved card before looking at whether fetching your usage worked
Unclear The exact difference you would see is not settled.
The message for an unsupported cloud-file pull now says it is not supported on this machine's platform, and no longer says "yet"
Memories Claude pulls into a conversation now have system-style tags rewritten so a memory file cannot pass itself off as an instruction
When file sync is off because you gave an explicit ref, the hint now says running without it starts from this checkout instead
A tool routed to another host is now refused outright when its permission check cannot be completed
Unclear Exactly which tools count as routed to another host.
Messages about sandbox violations now remove invisible and text-direction characters and every < and >
When several feedback surveys are waiting, Claude Code now picks one in a fixed order, preferring one that is actively asking
Unclear How surveys were chosen before, and whether each survey has its own switch.
Claude Code stops watching a settings folder for edits if the folder, or a link inside it, leads onto a network path
Plugin manifest parse and validation errors now end with the path of the manifest file, and are cut short later than before
When an installed plugin's version falls outside a required range, the error can now suggest updating that plugin
Unclear In which situations the hint is added is not known.
In a linked git worktree, or a checkout whose git folder lives elsewhere, Claude Code says file sync is off and the session starts from GitHub
Unclear It is not clear which command shows these messages, or whether a server-side setting controls them.
After a rejected token or a rate limit, Claude Code now stops re-asking the usage endpoint for a while, and retries once after a token refresh
Unclear It looks like the re-send after a token refresh runs even when the server-side setting is off, but this is not confirmed.
Claude Code now has a check that strips usernames, passwords and headers from plugin marketplace sources and refuses suspicious URLs
Unclear It is not clear when Claude Code runs this check, for example when saving or when displaying marketplace sources.
Claude Code's check for dangerous ssh, git and file addresses now ignores leading spaces and hidden tabs or line breaks before matching
Unclear It is not clear which Claude Code feature, such as plugin sources from git, relies on this check.
The permission prompt for reading a file outside your working folders now asks about this one read rather than reads in general
Unclear It is not clear whether what a Yes actually allows changed along with the wording.
A memory file's content now goes through the same cleaning step used when memories are written before it is added to Claude's context
Unclear What the cleaning step removes or changes is not clear.
In brief mode, loop updates are now sent as messages; the session-list tool now says when this session cannot send messages
Re-adding a marketplace under an existing name from a different source now says it was repointed, and how to undo that
Claude Code now joins backslash-continued lines before looking for ~ paths in shell commands, so a split path is still recognised
Claude Code checks fewer places for its Java proxy helper and no longer logs a warning when it is missing
Unclear It is not confirmed that the helper still logs anything at all in this build.
When a branch could not be restored, the error now says it may still name the session's HEAD, rather than stating that it does
Notifications built from several text pieces now use their own color if they set one, and fall back to warning color if they do not
Unclear The model and fast-mode messages may not set a color yet, so they may still appear in warning color.
Plugin manifest parse errors are now cut at 500 characters and validation errors at 1000
When Claude Code restarts itself from your home directory after you accepted the trust prompt, the restarted copy no longer asks again
Unclear Exactly which two paths are compared before the prompt is skipped is not confirmed.
Team cleanup now removes only the team folder and no longer deletes each team member's worktree
Unclear Worktree removal may have moved elsewhere rather than stopped.
A malformed image in a submitted prompt is replaced by a short text note saying it could not be processed, instead of causing an error
Unclear It is not known whether images arriving by other routes than prompt submission get the same check.
Published verbatim by Anthropic for v2.1.284. 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 100 bullets, 33 name something an entry on this page also names, 24 name something no entry here does, and 43 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-sonnet-5-5), now the default Sonnet model on the Anthropic API — 1M context, $2/$10 per Mtok with $0.20/Mtok cache reads
Probably refusal-fallback-now-has-its-own-route-for-claude-sonnet-5-5, sonnet-alias-now-resolves-to-sonnet-55-claude-sonnet-5-5, sonnet-55-claude-sonnet-5-5-added-and-made-the-default-s, model-constants-move-to-opus-55-and-sonnet-55 /usage and the status line (for example "$271.40 / $500.00 spent this month") when the gateway runs this version or later; the status line's rate_limits.spend_limit also gains used_usd, limit_usd and period
Probably gateway-docs-serve-get-apioauthusage-so-usage-and-the-s, usage-limit-auto-continue-text-is-remote-configurable-and-a, gateway-server-new-apioauthusage-endpoint-and-a-larger-h, self-hosted-gateway-serves-get-apioauthusage-from-its-spe, rate-limit-auto-continue-messages-now-come-from-remote-confi, statusline-and-usage-screen-show-spend-limit-dollars-for-gat, status-line-json-gateway-spend-limit-gains-used-usd-limit, max-20x-users-at-a-usage-limit-are-pointed-to-usage-credits effortSlider:decreaseEffort, increaseEffort and toggleUltracode keybinding actions, so the /effort slider's arrow and Tab keys can be rebound in keybindings.json
Probably effort-no-longer-preselects-ultracode, ultracode-becomes-an-onoff-row-in-the-effort-picker-instea, effort-slider-gets-leftright-keys-and-tab-to-toggle-ultraco /rate-limit-options to /help and the command menu for claude.ai subscribers, so the usage-limit notices that mention it point to a command you can find
Probably rate-limit-options-is-no-longer-hidden-and-has-a-new-descri /mcp reconnect all in the interactive terminal to retry every MCP server that failed to connect or needs authentication at once
Probably mcp-step-up-re-auth-tells-the-user-a-browser-is-opening, interactive-mcp-reconnect-all-in-the-terminal-ui, mcp-reconnect-all-now-reconnects-every-server-instead-of availableModels is empty, or leaves out the model Claude Code starts on without setting model or enforceAvailableModels
Probably gateway-config-validation-warns-about-empty-or-default-less auth: { google: {} } for Claude apps gateway telemetry.forward_to destinations, so telemetry can be exported straight to Google Cloud's OTLP endpoint using the gateway's Google Cloud credentials
Probably managed-gateway-config-validates-telemetryforward-to-auth, gateway-telemetry-forward-to-can-authenticate-with-a-google private_key_jwt) between the Claude apps gateway and its identity provider, for identity providers that issue certificate credentials instead of client secrets
Probably mcp-oauth-a-private-key-jwt-auth-method-is-registered-as-n, gateway-oidc-supports-private-key-jwt-certificate-client-aut source, and failing on every later turn after a malformed document block; a malformed image is now replaced with an explanatory note
Nothing to match on /usage, /extra-usage and IDE usage views now back off instead of re-asking
Probably gateway-docs-serve-get-apioauthusage-so-usage-and-the-s, usage-limit-auto-continue-text-is-remote-configurable-and-a, gateway-server-new-apioauthusage-endpoint-and-a-larger-h, self-hosted-gateway-serves-get-apioauthusage-from-its-spe, rate-limit-auto-continue-messages-now-come-from-remote-confi, statusline-and-usage-screen-show-spend-limit-dollars-for-gat, max-20x-users-at-a-usage-limit-are-pointed-to-usage-credits claude mcp add reporting success when managed settings restrict MCP servers to plugins; it now refuses and says what to do, instead of saving a server that never loads
Probably managed-policy-can-block-claude-mcp-add-when-strictplugino /plugin configure screen: boolean options are now a true/false choice instead of free text, number options refuse invalid input, and ←/→ change an options field instead of switching tabs
Probably marketplace-add-failure-now-raises-a-typed-error-carrying-th, invalid-plugin-manifests-reported-as-specific-errors, marketplace-name-conflict-error-now-carries-a-structured-kin ANTHROPIC_FOUNDRY_RESOURCE being interpolated into the Foundry endpoint host unvalidated; a value that is not a plain resource name is now refused
Probably anthropic-foundry-resource-is-validated-before-building-the, anthropic-foundry-resource-is-validated-and-passed-to-the-fo bin/ directories that don't exist are no longer added to PATH, and inherited entries aren't added twice
No entry names this sparsePaths plugin marketplaces cloning empty and replacing a working local copy on older git (before 2.39), which failed every refresh with "marketplace.json file is no longer present"
No entry names this [ in transcript mode writes the conversation to scrollback (macOS and Linux)
Nothing to match on /config and /plugin breaking the title and tab labels mid-word in a narrow terminal; a tab that doesn't fit now moves to the next line whole
Probably marketplace-add-failure-now-raises-a-typed-error-carrying-th, invalid-plugin-manifests-reported-as-specific-errors, turning-prompt-suggestions-back-on-in-config-resets-the-unu, marketplace-name-conflict-error-now-carries-a-structured-kin, new-unattended-commands-from-cloud-sessions-consent-row-in /model picker showing "+1 model" below the list after scrolling to the last model; the count now covers only the models below the visible rows
Probably gateway-config-validation-warns-about-empty-or-default-less, effort-migration-no-longer-special-cases-ultracode, sonnet-46-labelled-legacy-in-the-model-picker, model-picker-adds-sonnet-5-as-previous-sonnet-version, ultracode-becomes-an-onoff-row-in-the-effort-picker-instea, sonnet-alias-now-resolves-to-sonnet-55-claude-sonnet-5-5 /keybindings writing Backspace and Delete bindings for a footer action that does nothing into the generated keybindings.json
No entry names this footer:close) typing "x" instead of itself on the row of the agent you're viewing
No entry names this . not repeating text typed very fast (for example over ssh or in tmux) or pasted without bracketed paste, and leaving the prompt in INSERT mode after repeating a change with nothing typed (such as cw then Esc)
Nothing to match on dd on the last line or yy at the end of the prompt, where r or x would break or delete the image
Nothing to match on .claude/rules from outside the project being skipped without ever showing the external-imports approval prompt; a .claude directory symlinked from outside the project now asks for the same approval
Probably symlinked-clauderules-that-point-outside-the-project-now-n, external-import-warning-now-mentions-clauderules allowed-tools under managed allowManagedPermissionRulesOnly; only plugins from an official Anthropic source or a source that managed settings vouch for keep that pre-approval
Probably allowmanagedpermissionrulesonly-plugins-keep-allowed-tools, plugin-skill-allowed-tools-under-managed-only-policy-now-con, managed-permissions-policy-marketplace-npm-sources-must-pin, plugin-allowed-tools-trust-check-now-receives-plugin-provena claude plugin install leaving the plugin enabled and recorded when a dependency's version range could not be met
No entry names this {"decision":"block"} returned by Elicitation and ElicitationResult hooks being ignored; it now declines the MCP elicitation, as exit code 2 does
No entry names this SendMessage tool (such as by Claude Desktop) still being told to message other sessions with it
Probably peer-message-notice-drops-the-reply-via-sendmessage-hint-w, peer-session-messages-no-longer-tell-claude-to-reply-via-sen, loop-wakeup-guidance-is-aware-of-brief-mode-peer-list-tool /upgrade to users already on the highest Max plan; the warnings and /upgrade itself now point at /usage-credits when it is available
Probably usage-limit-auto-continue-text-is-remote-configurable-and-a, rate-limit-auto-continue-messages-now-come-from-remote-confi, max-20x-users-at-a-usage-limit-are-pointed-to-usage-credits /loop status updates in self-paced mode often not being shown because Claude wrote them only in its reasoning; Claude now writes each update, and the outcome when the loop stops, as visible text
Probably loop-self-pacing-prompt-reordered-stop-mentions-final-upda, loop-dynamic-pacing-prompt-reorders-confirm-and-arm-steps-p, schedulewakeup-loop-tool-results-now-tell-the-model-to-pos, loop-wakeup-guidance-is-aware-of-brief-mode-peer-list-tool, pr-steward-guidance-added-to-loop-and-pr-monitoring-prompts /ultrareview failing to upload the working tree when started from a git worktree that the Claude desktop app created on macOS or Linux
No entry names this data/users/ subtree, and added a "view" level to as_level
Probably artifact-write-db-if-version-now-required-on-writes-to-exis 431 Request Header Fields Too Large to every request from a sign-in whose identity provider lists many groups; it now accepts request headers up to 256 KiB
No entry names this /claude-api: hillclimb no longer spends rounds on prompt rewordings too small for the eval to measure, and an extra page you ask for beside report.html is built as one local file that loads nothing from the network
No entry names this /tasks, /copy and /hooks: the details after each name now line up in one column when they fit, and otherwise sit at the right edge
No entry names this claude plugin marketplace add to say when it replaces a marketplace already added under the same name from a different source, and how to undo it
Probably hidden-claude-plugin-marketplace-add-from-link-flag-for forceLoginMethod or forceLoginOrgUUID) and an API key, token or apiKeyHelper is configured: it now names the credential in use, where it is set, and how to remove it
Probably auto-mode-setup-checks-login-before-scanning-with-new-erro, detailed-error-when-managed-login-policy-conflicts-with-an-a MEMORY.md and recalled memory notes before they reach Claude
No entry names this claude remote-control: in a folder you haven't trusted yet, it now asks for workspace trust on the terminal instead of exiting
Probably claude-remote-control-help-notes-updated permissions.defaultMode still overrides it
No entry names this /effort (Tab, or /effort ultracode [on|off]): it no longer forces xhigh effort and stays on at any effort level
Probably effort-no-longer-preselects-ultracode, ultracode-becomes-an-onoff-row-in-the-effort-picker-instea ANTHROPIC_DEFAULT_OPUS_MODEL or modelOverrides: on the Anthropic API, the API now picks the model to switch to for each kind of flag, not the pinned model
Probably gateway-config-validation-warns-about-empty-or-default-less --allowedTools or an mcp_tool hook, even when CLAUDE_CODE_MCP_STARTUP_WAIT_MS is 0
Probably headless-mcp-pre-wait-now-tracks-servers-named-by-allow-rule, headless-first-turn-mcp-wait-now-holds-for-required-server /recap to decline with a short notice when it arrives relayed from a chat thread (your own included) or from a routine or webhook; typed in the terminal, the Claude apps, Remote Control, -p or an SDK host, it runs as before
Probably recap-refuses-requests-that-did-not-come-from-the-user-in-t /artifacts to show its filter tabs beside the title with one-word labels (All, Mine, Shared), using the same tab bar as /config and /plugin
Probably marketplace-add-failure-now-raises-a-typed-error-carrying-th, invalid-plugin-manifests-reported-as-specific-errors, turning-prompt-suggestions-back-on-in-config-resets-the-unu, marketplace-name-conflict-error-now-carries-a-structured-kin, new-unattended-commands-from-cloud-sessions-consent-row-in \\host\share path or a /net automount) unless it is on a mapped network drive added with --add-dir
Probably symlinked-clauderules-that-point-outside-the-project-now-n /model and Enter printing usage text into the chat instead of opening the model selector
Probably gateway-config-validation-warns-about-empty-or-default-less, effort-migration-no-longer-special-cases-ultracode, sonnet-46-labelled-legacy-in-the-model-picker, model-picker-adds-sonnet-5-as-previous-sonnet-version, ultracode-becomes-an-onoff-row-in-the-effort-picker-instea, sonnet-alias-now-resolves-to-sonnet-55-claude-sonnet-5-5 /feedback on Vertex, Bedrock and Foundry being refused after you pressed Send; the report is now saved on this computer, as the terminal does
Probably sdk-submitfeedback-can-save-a-feedback-report-locally-inste /mcp, /config or /settings command being shadowed by the extension's dialog
Probably mcp-step-up-re-auth-tells-the-user-a-browser-is-opening, turning-prompt-suggestions-back-on-in-config-resets-the-unu, new-unattended-commands-from-cloud-sessions-consent-row-in, interactive-mcp-reconnect-all-in-the-terminal-ui, mcp-reconnect-all-now-reconnects-every-server-instead-of CLAUDE_CONFIG_DIR in the claudeCode.environmentVariables setting to apply only when it is an absolute path, and passed it to terminals that continue the chat
No entry names this 1 added and 1 removed, of 218 lines, about 4 words, in the prompt 15 of 27 arms receive. 3 other prompts also changed. 1 of 27 tool descriptions changed. The appended system-reminder blocks moved: 1 line added.
Claude Code, interactive mode
153 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 79 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?