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 →Anything you can use today, anything that visibly changes, and anything worth poking at. One line each, open for detail.
Cloud sessions gain two new server gates, tengu_violin_maple for staying attached and tengu_violin_duet for directory sync
Unclear It is not clear what a connected session or file sync actually changes about the cloud session that gets created, or what the file sync switch falls back to by default.
Peer, coordinator and forwarded messages can be wrapped in id-tagged system reminders, and text that imitates a reminder tag is now escaped
Unclear What the reminder id is used for once it is attached is not known.
Behind a flag, /autofix-pr, /loop and PR-monitoring prompts now avoid pull requests labelled for PR Steward and offer the user choices
After over an hour idle, Claude Code can ask the server to clear old tool results, and file re-reads then skip the 'unchanged' shortcut
Unclear It is not traced where the mode setting is read, so the exact behaviour of shadow mode is known only from how the code checks it.
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
Unclear Model settings sent from the server may override these defaults.
Remote sessions can move a hash-verified opening of --append-system-prompt text into the cached system prompt via CLAUDE_CODE_APPEND_PROMPT_HEAD
Unclear It is not clear which cloud session launcher sets this variable, or whether any does yet.
A built-in cc-plugin-otlp can send OpenTelemetry records through plugin hooks, and plugin telemetry hooks must name a stream, with anthropic reserved
Unclear Whether plugin hook modules are available to everyone or sit behind their own switch is not clear.
The prompt to load mods, hook code Claude wrote, now puts "? How does this work" first and can explain up to 3 times
Unclear It is not clear what makes this question appear or whether a feature flag controls it.
Plugin hooks can now be passed to a separate plugin hook runner, possibly on another machine, with no visible way to switch it on yet
Unclear It is not clear whether anything sets up the outside runner yet, so this may not be in use.
The Remote Control dialog can now show an extra row for unattended serving, with a change action and an optional notice
Unclear What decides whether the row appears, and what its text says, is not known.
In cloud sessions, a tool that may write can now wait up to 30 seconds for its call to be stored, if the server's upload policy asks for it
Unclear It is not clear whether any server sends these instructions yet, or whether 30 seconds is a default or an upper limit.
Usage-limit grace-window text gains per-model and subagent-specific remote settings, and the near-limit notice now counts as a wrap-up note
Unclear It is not clear what this text says or when Claude Code inserts it.
A 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
Unclear It is not clear when the row appears for someone who has never answered the consent question.
SDK 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
Unclear The request is marked internal, so it is unclear which apps actually send save_locally.
Artifact read_db and write_db add a view access level, require if_version on existing documents, cap batches at 1 MiB and explain who can read what
The sandbox now refuses allowed hostnames that resolve to loopback, cloud-metadata and similar addresses, and deniedResolvedAddresses adds your own ranges
Unclear Whether this setting can be set in the sandbox section of settings.json, rather than only inside the sandbox's own configuration, is not confirmed.
A command for switching account memory and a permission dialog for writing to it are listed in the code but left out of this build
Claude Code now carries new rules for retrying failed API calls, which run alongside the current ones only to compare decisions
Unclear It is not confirmed whether any part of Claude Code acts on the new rules' decisions or whether they only observe.
Footer labels for account memory being read-only or off are in the code but switched out, so nothing new shows
Auto mode's action checker can use a stricter wording covering connected apps and accounts; CLAUDE_CODE_AUTO_MODE_CANDIDATE_WORDING selects it
Claude Code has checks for a cloud session with no disk, but nothing in this build turns that mode on
Unclear Another program hosting Claude Code, such as the desktop app, might be able to turn this mode on in a way that was not traced.
A new check can refuse or clean memory files containing invisible characters or tag-like text; it falls back to off unless enabled remotely
Claude Code's requests can now include an instruction to clear old tool results from context, not only old thinking
Unclear It is not clear what supplies the tool-clearing instruction or whether a flag controls it.
Gateway telemetry.forward_to entries accept auth.google for Google Cloud endpoints, and gateway models now advertise 1M-context support
For gateway logins with a spend limit, the status line data and usage screen now include dollars used, the dollar limit and the period
Ultracode is no longer an effort level: it is a session toggle in /effort, with new SDK fields and a narrower subagent-cap exemption
Unclear It is not settled whether the ultracode row is hidden on plans where workflows are off by default.
The Agent SDK now ends a run when Claude Code reports it is idle, which can mean waiting up to 10 minutes after the result
Unclear Exactly what ending the run shuts down in each input mode, single-turn versus streaming, is not settled.
/mcp reconnect all reconnects every MCP server#/mcp reconnect all now reconnects every MCP server that needs it instead of looking for a server named "all"
Unclear It is unclear whether the interactive view already accepted reconnect all another way before.
--model help text no longer gives an example full model name#The help text for --model now ends with "or a model's full name." and no longer shows an example full model name
The request for extended usage past a limit can now apply to the main conversation, not only subagents, but stays off unless switched on remotely
Unclear What the new condition that lets the main conversation through represents is not known.
Claude Code includes a new error saying Claude Design cannot be signed in from inside a Claude Code session
Unclear Where Claude Code actually shows this new message could not be found.
The Read and NotebookEdit tools can now accept results produced by another process, and tools aimed at an unknown machine or another session are refused
Unclear It is not clear what switches on sending tool calls to another machine, or whether any reader can use it yet.
When the product-feedback policy changes, Claude Code now re-sends every feature switch to the VS Code extension, not just two of them
The gateway's OIDC config accepts private_key_jwt with a certificate in oidc.client_assertion, so oidc.client_secret is now optional
The plugin runtime's model.complete accepts an optional abort signal and returns an "aborted" result instead of throwing an error
On the highest Max plan, the usage-limit hint and /upgrade now point to /usage-credits, since there is no higher plan to move to
In the effort slider, left and right change effort and Tab toggles Ultracode, and these keys can be rebound
Unclear It is not clear whether the Tab hint is shown only when Ultracode is available.
claude -p get a local command's exit code#In -p mode with text or json output, Claude Code now exits with the exit code a local slash command reported, instead of always 0
Unclear It is unclear which local commands actually report an exit code.
What would fix it?
Smaller changes and internals, grouped as the pipeline found them. Nothing is dropped, it is only further down.
9 more of these are in What probably matters to you, above.
claude --cloud file sync can now be paused from Anthropic's side without ending the session, and explains more precisely what gets uploaded
Unclear It is not clear what else decides whether cloud file sync is available at all in this build.
task#The enable tools for Claude in Chrome, the desktop app's browser and computer use now tell Claude when to call them and take an optional task
Unclear It is not settled when this tool is available or what calling it actually does beyond its new description.
Gateway addresses that fail the local checks can now be accepted if a recognized Claude gateway deployment vouches for them, using the session's login token
Unclear It is not clear which login steps use this confirmation or whether a switch controls it.
In Projects sessions, 'always allow' for WebFetch or WebSearch is saved for the session only, and hooks and plugins can no longer grant these tools
Unclear Apart from the environment variable, how Claude Code decides a session is a Projects session is not stated.
File sync for claude --cloud sessions now signs uploads with the bound device's key and shows a clear message when Anthropic has switched sync off
Unclear Whether Anthropic's server sends this refusal today cannot be told from Claude Code itself.
Remote Control now reports a reason when it is refused, and gains a skipAutoTitle option tied to disabled terminal titles
Unclear The wording of the reason, and which parts of Claude Code ask for it to be reported, are not clear.
Server-reported context_management results are kept on assistant messages, and auto-compact adds the cleared tokens to its estimate
Unclear It is not clear why the cleared tokens are added to the estimate rather than taken away from it.
When you hit a usage limit, Claude Code shows a live status and next step, suggests /usage-credits, and the wording can come from remote config
Unclear It is not clear where on screen the new status is shown.
The conversation view can now fold agent messages, and the "N new messages" divider moves to the next visible row when its message is folded
Unclear It is not clear which setting or view mode switches agent message folding on.
On macOS with the sandbox or a proxy on, project hooks run through remote tools can now run confined instead of being refused, behind a flag
Unclear It is not clear what the second condition checks or which parts of Claude Code reach this hook runner.
When a plugin copies text outside the terminal, Claude Code now hands it to a connected client that supports copying instead of giving up
Unclear It is not clear which client apps support these copy and prompt actions.
/memory shows an Account memory status line#The /memory screen shows an "Account memory" status line, with a note, below the synced project memory line
Unclear What fills in the Account memory line, and when it is shown, is not known.
The model picker now describes Sonnet 4.6 as "Sonnet 4.6 · Legacy" instead of the previous Sonnet version
In diskless sessions, built-in file and shell tools that cannot reach the cloud environment now stop with a clear 'retrying will not help' error
Unclear Which tools are in the set, and which kinds of session count as diskless, is not clear.
Cloud-session status can now say a feature is switched off by Anthropic and rechecked on every message, and gains a device_unbound reason
Unclear Where the new text is shown, and what situation device_unbound describes, is not known.
Switching auto memory in /memory or in settings now also writes an account-related value to your user settings
Unclear What the account-related value holds and what it is used for is not clear.
When a cloud session attaches, Claude Code now offers directory sync without first checking a remotely controlled switch that could hold it back
Unclear It is not settled whether another check earlier in the attach process still decides whether sync is offered.
When a model is blocked, the error now shows the block message sent by the server, or a generic line with a link to Anthropic news
Unclear Where the server's explanation comes from is not stated.
In remote containers, directory-sync snapshots now fire after 5 writing tool calls instead of following quiet-time and max-wait timers
Unclear Which sessions use this folder sync, and what turns it on, is not known.
When a remote host misses check-ins but still answers a liveness probe, Claude is told the reply was lost and to try a narrower command
Unclear It is not clear which remote host feature this applies to or whether a flag controls it.
The bundled gateway guide now tells operators how to serve spending data so /usage and the status line can show dollars
Unclear The client-side request to this endpoint was not confirmed in this build.
/recap now refuses requests relayed from Slack, Teams, project threads, routines or other programs, and only answers you directly
The gateway bundled with Claude Code now answers usage requests when spending is tracked, and allows request headers up to 262144 bytes
Unclear What sets up the spend metering is not known.
opusplan model setting now gets a much shorter system prompt#With opusplan, Claude Code now sends Claude a compact system prompt with a short Harness section and a condensed memory guide
Unclear Why opusplan switched to the short prompt is unclear; it may be because it now resolves to a different underlying model.
Sandbox path patterns now match through linked folders, and denyRead rules are no longer undercut by default writable folders
Unclear Where the new IP address range handling is used is not confirmed.
A self-hosted gateway with spend limits now answers usage requests with each user's caps, so clients behind it can show remaining usage
The gateway can now turn its spending caps into a usage summary in US dollars, with warning and critical levels
Unclear Where this summary is shown, and whether it is switched on, is not clear.
The prompt for reading files outside your working directories gains an allow-once answer that permits one read and asks again next time
Unclear It is not settled whether this prompt opens on every first outside read in auto mode or only under narrower conditions.
Auto mode's safety check has a second version of its user-boundary rule that also covers committing changes in a connected app or account
Unclear It is not clear which part of Claude Code uses the version that covers connected apps.
The narration text that could appear under the working spinner is gone, while tips and the "Next:" task line still show
Unclear It is unclear whether the narration line was ever shown to users, or whether it now appears somewhere else.
Claude Code can show your plan as "Claude" plus the name your organization's account gives, instead of a fixed label
Unclear It is not confirmed which screen shows this subscription name.
When SendMessage is missing, the note on an incoming peer message just says to decide whether to respond, not to reply via SendMessage
Unclear Which session setups leave the SendMessage tool unavailable is not stated.
A proposed new skill that shares a name with one of your own skills is refused unless Claude has read that skill's SKILL.md in full
Unclear Whether the skill-proposing tool itself is only available to some accounts is not stated.
Syncing a folder's files to a cloud session now stops, with a message, when this computer can't sign uploads as the device tied to the session
Unclear Whether this path is behind a switch, and what that switch defaults to, is not stated.
The warning about thinking from another organisation now says it is not deleted and comes back if you resume signed in to that organisation
Unclear Whether Claude Code now actually keeps the thinking, or only the wording changed, is not settled.
Sessions launched as diskless now refuse claude.ai marketplaces and mods with a reason, skip ripgrep, and stop pointing the model at a transcript file
Unclear Which hosts launch sessions in this mode is not known.
New instructions tell Claude that pages it generates must work at about 400px wide and never scroll sideways
Unclear It is not clear when these instructions are given to Claude.
Claude Code now explains why file sync is off for a cloud session on your computer, such as the session not being tied to this device
Claude now tells you once when Anthropic has switched off directory file sync, and again when it is back on, checking at the start of each turn
Unclear It is not clear what causes Anthropic to switch file sync off.
Branch sync from a cloud session now undoes or refuses a branch move if git's records changed underneath it, and rejects unsafe Windows file names
Unclear It is not clear what switches this sync feature on.
Cloud folder sync refuses to start without a device signing key, and any request it cannot sign is stopped rather than sent unsigned
Unclear The settings or switches that make cloud folder sync available in the first place are not described.
Automatically written memory and pinned memory now have hidden characters cleaned out and tag-like text escaped before they reach Claude
Unclear Exactly which characters are removed is not settled.
When a remote tool answer is too big, Claude Code now shrinks embedded images to fit before refusing, and the size limit can be set remotely
Unclear The lowest and highest values the size limit is held between were not checked.
Starting a cloud session that would sync a local folder now binds this device first, and drops the sync with a notice if that fails
Unclear It is not clear which remote switches decide when connected sessions and file sync are used.
When you start a cloud session, the summary now names tracked files that must stay local and says the session runs without them
Unclear It is unclear which files count as ones that must stay on this computer.
Cloud file sync now signs synced files with a key for this computer, and stops with a clear message if that cannot be set up
Unclear Whether this is controlled by the remote switch it appears to be is not confirmed.
Pausing or resuming memory makes Claude Code re-read your memory and instruction files right away, with a note saying why
Unclear It is not clear whether pausing memory previously left out-of-date memory in Claude's context by some other route.
Plan mode's "Cannot call this while in plan mode" prompt now covers more tools, and a whole-tool allow rule no longer approves some computer-use actions
Unclear It is not clear what makes a tool count as having a session-level setting of its own.
When a cloud session claims auto mode but your computer checks a permission in another mode, the prompt now says why
Unclear What decides that the computer is not accepting auto mode from cloud sessions.
The Claude session link on commits and pull requests is now added whenever a session id is found, not only in remote mode
Unclear Where the new session id comes from, and so which sessions are affected.
When ultracode is unavailable, the error now names one reason, and turning it on no longer forces extra-high effort
Unclear It is not clear whether effort is now set for ultracode somewhere else.
Upload-only folder sync can now bind to your machine, and a folder cloud session that cannot bind is archived instead of created
Unclear It is not established what controls whether this folder sync feature is available at all.
Long messages you send now show only their first lines in the conversation, with a count of hidden lines you can expand
Unclear How many lines are shown before shortening, and whether some modes turn this off, is not known.
claude mcp add is refused when managed settings allow only plugin MCP servers#If managed settings restrict MCP servers to plugins, claude mcp add now stops with an error instead of adding a server that never loads
Starting claude rc in an untrusted folder now shows a trust prompt with a y/N question instead of printing an error and quitting
Unclear Whether anything besides both input and output being a terminal decides if the prompt can be shown.
A successful Bedrock or Vertex model-access check is now reused for 15 minutes, keyed on your region, endpoint and credentials
Unclear Only the Bedrock path was traced end to end, and the 15-minute window was not confirmed to belong to this cache.
When unsandboxed commands are turned off, Bash and PowerShell no longer offer Claude the dangerouslyDisableSandbox option
Unclear It is not clear which way of running Claude Code turns unsandboxed commands off in this way.
Masked secret files that can't be masked fall back to a read deny, and denyRead now also shapes what sandboxed commands may write
Unclear Where the additional paths come from is not known.
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
71 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?