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 →When a session ends, Claude Code now closes only the tmux or iTerm teammate panes that session opened, not every one in the team
If Claude calls an MCP tool before its server has finished connecting, the call is queued rather than rejected as an unknown tool
Unclear How long a queued call waits for its server is not settled.
When the service refuses a tool's answer in a Remote Control session, Claude Code now sends an error result instead of dropping it
Unclear When exactly this replacement is triggered is not settled.
The usage-limit dialog now takes the reset time for its auto-continue option from a snapshot made when the dialog opens
Unclear Exactly how the new calculation of the reset time differs from the old one in practice is not settled.
proxyAuthHelper is now sent on more connections#The proxy sign-in header produced by proxyAuthHelper is now also attached to connections made through Claude Code's other proxy route
Unclear It is not clear which of Claude Code's requests go through the route that now gets the header.
/effort no longer starts on ultracode when you open it#The /effort picker no longer jumps to ultracode as its first selection when ultracode is on
Unclear The visible difference in the picker is not confirmed.
An Elicitation hook that signals a block now always blocks, even if it also reports success or its output is empty or not JSON
Mouse clicks and terminal focus changes now wait for pending screen updates to finish first, as key presses already did
When a fullscreen view closes, Claude Code now restores and clears the normal terminal screen instead of just redrawing over it
Unclear Which fullscreen views actually use the new clean-up is not known.
Your sandbox filesystem.denyWrite paths are now enforced in the sandbox mode used when a proxy port is configured, where they were ignored
Unclear What this particular sandbox mode is used for is not clear; it may be a sandbox for scripts or checks.
When a saved conversation is resumed, Claude Code now drops invalid turn-position values and strips producer details from task notifications
On Linux, a sandbox setup too long for one command line is now passed to bubblewrap through a temporary file, with clear errors when it can't be
Switching prompt suggestions back on in /config clears the count of ignored suggestions, so they resume at once
Removing a permission rule or editing a hook is now refused only when the settings file has real errors, not just warnings
Unclear Whether the hook-editing check is also changed in this release, or only the rule removal.
Pulling a queued message back into the input to edit it now keeps its pasted images as placeholders in the text
Unclear This behaviour was inferred without seeing how the placeholders are written.
After a fullscreen view closes, Claude Code rebuilds the normal screen from a blank frame, and a text selection now forces a full repaint
Unclear Which fullscreen views this affects and what visible problem it fixes are not stated.
A plugin dependency given by name alone is now satisfied by, and resolves to, the matching enabled plugin from a marketplace
Unclear It is not clear exactly how plugin names are rewritten into a consistent form before they are compared.
Directory sync drops files that are gone from disk, or match an exclusion, from its list of withheld files
Unclear It is not clear which paths the exclusion covers.
subagentStatusLine command is skipped when the status line is switched off#Claude Code no longer runs your subagentStatusLine command when its restricted-mode check has turned off the statusLine feature
Unclear It is not certain that the check that now skips the command is exactly the one --bare mode uses.
In vim mode, repeating a change that inserted no text now leaves the cursor one character back, unless it is at the start of a line
Unclear It is not clear which keystrokes produce a repeated change with no inserted text.
A link check now removes tabs and line breaks anywhere in a link before reading its scheme, such as http or javascript
Unclear Which feature relies on this link check is not settled.
A local MCP server that rejects start-up after the version check timed out is now restarted once without that check
Two GitHub plugin or marketplace sources whose repository names differ only in capital letters are now treated as the same source
Messages about remote approval requests that timed out, ended or were withdrawn now fill in their actual timeout, tool name and reason
Unclear It is not settled whether the old messages actually showed wrong values or whether this was only a rename inside the code.
Accepting the suggestion to switch to medium effort now takes your current effort and ultracode state into account
Unclear What difference this makes in practice is not clear.
A broken piece of a streamed reply now stops with the error "Stream event unreadable" instead of carrying on with bad data
Unclear Whether Claude Code retries the request after this error is not clear.
Claude Code no longer offers an API error message as a suggested next prompt
In vim mode, a whole-line command now starts at the beginning of the line when the cursor sits just after a pasted-text placeholder
Unclear Which kinds of placeholder besides pasted text count is not clear.
When you switch Remote Control's spawn mode, the usage report Claude Code sends now records the mode you switched to
Unclear It is unclear whether the report actually records a different value now, or whether the code only renamed the value it already recorded.
When no tool.call hook answers a plugin-served MCP tool, the hint now builds the tool name from the MCP server's name, not the plugin's
/reload-plugins really does#When you decline reloading a mod, Claude is told it loads at the next session start and that /reload-plugins only reloads mods loaded at start
enabledPlugins are matched regardless of letter case#An enabledPlugins entry for a built-in plugin now takes effect even if its name differs from the plugin's only in upper or lower case
Unclear The case-insensitive matching was inferred rather than confirmed.
In automatic protocol mode, Claude Code now retries an MCP server with the newer protocol even when its first refusal lists no versions
In vim mode, yanking to the end of the text moves the cursor differently, and dot-repeat stores inserted text in one standard character form
Unclear Where exactly the cursor now ends up after yanking to the end of the text is not stated.
Claude Code no longer adds a folder to PATH a second time when it is already there for Bash commands
16 more of these are in What probably matters to you, on page 1.
Behind server-side switches, computer-use actions can skip allow-rule shortcuts in auto mode, and browser and desktop tools can ask for a short action summary
Unclear It is not clear what reviews these actions once they go past the allow rule, whether an automatic classifier or a prompt to you.
Claude Code-owned MCP configs can now register a provider that adds headers to each request, but nothing in this build registers one
Unclear It is not clear which future feature will use this or exactly which MCP connections it will apply to.
Remote sessions started from the Claude desktop app now depend on the same single flag as every other starting point
Unclear It is not clear which feature tengu_violin_wood itself controls.
Attaching served remote tools to a session now checks a remote switch first and reports "flag_off" instead of loading the serving code
Unclear It is not fully confirmed which remote switch this check reads.
In sessions started by a remote trigger, one tool is no longer always forced into the set of tools loaded at the start
Unclear Which tool this applies to, and whether it was always loaded up front before, is not settled.
Claude Code's check on whether a command is available now consults an extra per-environment switch that nothing turns on
Unclear What will eventually set this switch, and which commands it would hide, is not known.
Permission prompts for writing Markdown memory files can now show a header, but its text is empty so nothing appears yet
How the header shows your working folder in remote mode can now also be switched by the remote setting behind connected cloud sessions
The telemetry event format adds fields for the git remote URL and working-tree state at session start, but nothing fills them yet
5 more of these are in What probably matters to you, on page 1.
Agent tool results and hand-back recipients now receive the subagent's untruncated report, with any auto-mode security warning
Unclear It is not clear whether the SubagentHandback tool is available to everyone or switched on only for some accounts.
When tool-result clearing is active, the thread planner sends the turn as a fresh stateless request instead of continuing a server-side thread
Saved pause records for transcript sync can now store an account-level choice of on, read-only or off, with two new pause reasons
Unclear It is not clear which sync feature these account choices control or where a reader would set them.
After fetching skills from the server, Claude Code records which ones it owns, and only logs a warning if that record cannot be saved
Unclear It is not clear what Claude Code will use the ownership record for.
Promo eligibility is now saved per account and endpoint and reused for up to an hour, with a new statusMaxAgeSeconds setting controlling the age
When publishing an artifact, Claude Code now saves the list of artifact capabilities to your config as cachedArtifactRoster
A new readBack option asks the server to read state back for artifact frame requests and list_files actions, with read_back noted on share status
Unclear Which tool or connector this belongs to is not known.
On first-party requests, x-anthropic-billing-header now includes cc_prompt_index and cc_turn_index when both are known
Unclear One additional condition on when the header is sent was not identified.
A gateway host can now be accepted because a recognized deployment vouches for it, and those hosts are not pinned to a network fingerprint
Unclear It is not clear what uses this gateway check or how a deployment comes to vouch for a host.
Remote Control sessions normally get an automatic title from the 1st and 3rd message, and under some condition this is now skipped
Unclear What triggers the skip is not clear; it could be a setting, a switch, or a title you already set.
File edits whose input came from another machine now get the same path and text clean-up that remote edit calls already got
Unclear It is not clear which kinds of session mark input as coming from another machine.
Stored MCP OAuth credentials now remember which auth server and metadata URL were used, and keep them across refreshes and invalidation
Unclear The effect on login refresh is inferred rather than shown.
The ListAgents tool's description is now built according to whether the agent can send messages to other agents
Unclear The wording of the two versions of the guidance is not known.
Queued notifications relayed from another session are now dropped when their id is longer than a length limit
Unclear The length limit itself.
Parts of the terminal display can now update their hover highlight when content redraws under the mouse, not only when the mouse moves
Unclear It is not clear which parts of the interface use this.
Claude Code now records timing checkpoints when idle-time tool-result clearing starts and when a computer-use action is first summarised
Unclear What controls whether those two features themselves are switched on is not known.
When a project pre-tool hook cannot be sandboxed for a remotely served tool call, Claude Code now records why; the hook's outcome is unchanged
Unclear It is not clear how remote tool serving is switched on.
Some usage data now carries a request ID and a prompt ID, and logged request errors now include the request headers
Unclear It is not clear which records gain the IDs or where they are sent.
Claude Code now logs when an ElicitationResult hook changes the answer you gave an MCP server, or when a hook answered for you
Claude Code now logs when it starts showing fewer prompt suggestions and when that ends, with no change to what you see
Claude Code's API error reports now record whether it skipped its non-streaming fallback, a retry path for failed streamed replies
Unclear What causes the fallback to be skipped is not clear.
Directory-sync publishing stops waiting for a snapshot and drops its hold-time figure; sandbox fallback reports now include the sandbox mode
Unclear Whether anything relied on the directory-sync step waiting for the snapshot is not known.
Each SDK result report now records the turn number and how many MCP servers were still connecting when the turn started and ended
On shutdown, one telemetry sender now waits up to half the flush time limit for pending work to finish before sending
Unclear Which telemetry sender this is, and what work it waits for, is not clear.
Claude Code no longer tells Claude to reuse an earlier fetch of a page once that earlier result has been cleared on the server side
Unclear When the repeat-fetch note is shown at all is not settled.
Task notifications about work the session launched itself, like background agents or shells, can now be labelled as such
Unclear What sets the label is not known.
Saved cost data is now rejected when its token totals across models are larger than the limit allowed for a single number
Unclear It is not stated what happens to the saved cost data when it fails this check, or what the limit is.
The WebSearch tool's description now says the month is provided in the conversation instead of naming a month and year
Unclear It is not settled which paths still fill in the real month, or whether every session gives Claude the date somewhere else in the conversation.
When Claude Code registers as an OAuth client, a configured private_key_jwt authentication method is now sent as "none"
Unclear Whether this applies to MCP server sign-in specifically is not confirmed.
Claude's web search instructions no longer name September 2026 as the current month and point to the date given in the conversation
A debug log message about the rules folder being a symlink now also refers to the .claude folder that contains it
Unclear Only the log wording is known to have changed. It is not confirmed whether Claude Code now actually treats a symlinked .claude folder like a symlinked rules folder.
Claude Code now accepts an embedded image only if its data is actually text, and handles damaged content without crashing
If a plugin's stored data file is missing, Claude Code now tries to copy it over from a related session before starting empty
Unclear It is not clear which source the plugin data is copied from.
When Claude Code finds an item by plugin name, it now matches on the name combined with the plugin's source, not the bare name
Unclear It is not clear what kind of plugin item is looked up this way or exactly how the combined name is formed.
When /plugin marketplace add fails to save a marketplace, the error now includes its name and folder for Claude Code to act on
Unclear Nothing is known to use the new detail yet, so whether any recovery is offered is not settled.
Re-adding a plugin marketplace with a different source now logs both the source it had and the one replacing it
Unclear It is not clear whether the replaced source is shown anywhere on screen or only in the log.
A refused /plugin marketplace add now records whether the marketplace is already added from the same address or its name is taken by another
Unclear It is not clear whether this kind is shown to you or only used inside Claude Code.
Checking a plugin hooks module now lists what it declares for usage data and can report problems for that module
Unclear What the checks look for, and what counts as a problem, is not clear.
In cloud sessions, message types the server names now get a temporary copy for live viewers while the saved version is held back
Unclear Which message types the server lists for this is not known.
Whether a session's file sync still owes its first upload now depends on Anthropic switching file sync off, not on the sync being muted
Unclear What muting a sync means for the user, and where this check shows up, is not settled.
A waiting status tied to device uploads can now report a failed git step or uploads still pending from an earlier run
Unclear Which feature or screen shows this status, and who sees it, is not known.
Git-based folder sync for cloud sessions now records extra diagnostics when a turn begins before the device has reported its status
Unclear It is not clear whether this case also shows you a message.
When Claude Code cannot tell whether two paths are the same file, the result is now reported as indeterminate instead of not live
Unclear What this live or not-live check protects, possibly git worktrees, is not settled.
A shared check for operations that take a path now only confirms a path was given and no longer refuses network locations
Unclear It is not clear which feature these operations belong to, or whether network locations are now refused somewhere else instead.
A directory listing inside Claude Code now returns each file's last-modified time alongside its size
Unclear Which feature uses this listing is not known.
When identifying a file by its device and inode numbers, Claude Code now ignores more placeholder inode values, including -1 and 2^64
Unclear Whether this applies only on Windows, and which features rely on this check, is not known.
On macOS, the check for whether a sandboxed feature can run now returns one of several modes instead of always refusing when a sandbox is already active
Unclear It is not clear which feature relies on this check, or what the new capability check behind the person mode tests.
Claude Code now lists an error type for Linux sandbox profile failures alongside its other named error types
Unclear What this list of error names is used for is not clear, and neither is whether the error type itself is new.
npm root -g on the spot#The step that looks for the sandbox runtime package no longer runs npm root -g itself and waits up to 5 seconds for it
Unclear Where the npm location now comes from, and so whether the pause is gone or only moved, is not known.
On Linux, sandbox setup now waits for a background search for the apply-seccomp helper unless seccomp.argv0 is set
Unclear Whether the set of folders searched for the helper actually changed is not known.
Shell command output in a remote session can now be cut to a given number of characters, and otherwise is shortened as before
Unclear Nothing in this build appears to supply the limit yet, so the cap may have no effect for now.
When a remote session fails to send an event because it is too large, that failure is now reported as its own case
Unclear Whether Claude Code does anything different when an event is too large is not settled.
Sending session events now reports when the service rejected them as too large, instead of treating it as a general failure
Unclear What remote sessions do after an event is flagged as too large, and whether it is retried, is not clear.
The request that fetches a remote session's settings now sends a User-Agent header
Claude Code now records how long startup waits for code to load before it takes over a ready-made spare process
Claude Code now records how long it takes to take over a pre-started spare process, with no change to what you see
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?