Follow Discord

Claude Code v2.1.286

348 entries read Grouped into 290 v2.1.285 → v2.1.286 Mods API: +9 added · 0 removed · 4 changed · 2 docs only Markdown JSON Follow Unofficial

You can now rely on a built-in security plugin to keep your deny rules in force when other plugins try to lift them. It can also refuse plugins that did not come from your organization. Plugins can now build the system prompt themselves through a new prompt.compose hook, and Claude Code falls back to its own prompt if that fails. Setting CLAUDE_CODE_AUTO_MODE_TIER to any value, even "full", puts auto mode on its strict tier. With CLAUDE_CODE_ARTIFACT_SHARE set, Claude can share an artifact with your organisation or named colleagues after you confirm. A new notice offers a key, ctrl+y by default, to make a newer model your default.

This release has 16 entries in the build that are not switched on yet. On Linux, project hooks can run in a locked-down sandbox called bubblewrap, with no network and a hidden home folder. That sandbox waits on a server switch. Keepalive pings that hold a one-hour prompt cache open during idle pauses are also off unless switched on remotely. A background task for importing local memory notes is taking shape, with a "memory import" entry in the task list. Remote tool permission checks gain retries for timed-out safety checks, behind three switches.

Prompts rewritten by a hook on submit now reach Claude as rewritten. Resumed sessions restart interrupted turns by mistake less often. Resuming in --print mode with an SDK connection now always runs its cleanup step, even after a failure. The ultrareview details dialog no longer offers to open the review in a browser. Starting a remote session no longer sends file sync, folder seeding or sync consent options. The /plugin-types command is now hidden from command listings but still works.

Written by our agent from the shipped bundle, not by Anthropic.

Five to try today

Picked from 22 you can use now
  1. New CLAUDE_CODE_AUTO_MODE_TIER variable selects the strict auto mode tier

    Setting CLAUDE_CODE_AUTO_MODE_TIER to any value selects the "strict" auto mode tier, and it is passed on to processes Claude Code starts

    Sessions & agents
  2. Built-in claude-test plugin gets clearer start-up errors and no longer requires a terminal session

    The claude-test plugin now explains why it failed to start, gives new advice on duplicate copies, and drops its terminal-only message

    Elsewhere
  3. Early plugin sync at startup for remote sessions, behind an environment variable

    A new early plugin sync step at startup runs only when CLAUDE_CODE_CCR_EARLY_PLUGINS_SYNC is set, and not in interactive sessions or with a proxy auth helper

    Extensions
  4. A new own_record option for unattended permission handling

    A permission setting now accepts own_record for unattended sessions, and one more outcome now counts as stopped

    Permissions & safety
  5. Built-in security plugin keeps deny rules in force over other plugins

    A built-in plugin restores deny rules that other plugins lift, and can refuse plugins that did not come from your organization

    Permissions & safety

Want the reasoning? Read walks the 44 entries that probably matter to you, each one opening to what changed and why.

Read this release → Every row →
Reading as
Show only
Tier
Flag state
Names
Pick an entry · j / k steps through · rest on a row to peek
124 entries

Improvementsopen

Continued from page 1

6 more of these are in What probably matters to you, on page 1.

↑Improved
You'll notice
Useful3 Signal1
Themes unclear

Theme picker previews as you move and restores your theme on cancel#

The theme picker previews each theme as you move through it, ticks the current one and restores your saved theme if you cancel

Unclear How many rows the output style picker now shows is not settled.

Details 0 0 Feedback
↑Improved
You'll notice
Useful3 Signal1
Model Fallback unclear

Switching to a fallback model now warns when the context window shrinks#

When Claude Code falls back to a model with a smaller context window, the notice shows both sizes and suggests /clear or a larger fallback

Unclear It is not clear whether this notice is shown to everyone or only when some setting allows it.

Details 0 0 Feedback
↑Improved
You'll notice
Useful3 Signal1
Input unclear

External editor now opens at the line your cursor was on#

When you open your prompt in an external editor, Claude Code now tells the editor to start at the line your cursor was on

Unclear Which editors are given the line number in this form.

Details 0 0 Feedback
↑Improved
You'll notice
Useful1 Signal2
MCP unclear

MCP tool calls can skip automatic backgrounding in some sessions#

Under a new condition, an MCP tool call no longer gets a timer for moving itself to the background, and plugin sync logs two more details

Unclear It is not clear which sessions meet the condition that turns off automatic backgrounding.

Details 0 0 Feedback
↑Improved
You'll notice
Useful1 Signal2
MCP no docs found unclear

Automatic MCP server discovery has one more condition that can switch it off#

Automatic discovery of MCP servers now also stays off when an additional check at the start of its conditions is true

Unclear What the new first check tests is not known, so it is unclear when it turns automatic MCP discovery off.

Details 0 0 Feedback
mcpAutoDiscovered
↑Improved
You'll notice
Useful1 Signal2
Elsewhere unclear

Questions without a stated type are treated as multiple choice#

When a feature check passes, questions Claude asks you that have no type are now marked as multiple choice

Unclear It is not clear what the feature check tests or who it is on for.

Details 0 0 Feedback
↑Improved
You'll notice
Useful1 Signal2
Artifacts unclear

Errors now explain when an option offered earlier is no longer available#

If Claude uses an option that was offered earlier in the conversation but is not available now, the error says so and tells it to continue

Unclear It is unclear whether the Artifact family of tools is itself switched on for everyone.

Details 0 0 Feedback
↑Improved
You'll notice
Useful1 Signal2
Sandbox unclear

More sandbox settings now block serving device tools remotely#

Claude Code now refuses to serve device tools when seccomp is unavailable or more sandbox-weakening settings are on

Unclear It is not confirmed which remote feature uses these device tools.

Details 0 0 Feedback
↑Improved
You'll notice
Useful2 Signal1
Remote Control unclear

Remote messages with files can start before the files are saved#

A remote user message with attached files can now start Claude's reply before the files are written, and Claude is told if any could not be saved

Unclear It is not clear what decides whether a message takes this earlier-start path.

Details 0 0 Feedback
↑Improved
You'll notice
Useful2 Signal1
MCP unclear

Retrying an MCP sign-in reuses the link that is already open#

If an MCP server sign-in is still in progress, Claude Code now reuses its link instead of starting a new sign-in that would break the first

Unclear How recent the earlier sign-in must be for its link to be reused is not stated.

Details 0 0 Feedback
↑Improved
You'll notice
Useful2 Signal1
Permissions

Bash permission prompt shows the description first#

The prompt asking permission to run a shell command now shows its description above the command and lists network hosts only when there are some

Details 0 0 Feedback
↑Improved
You'll notice
Useful2 Signal1
API Requests unclear

Retries of a failed response stream now fit inside the time left#

When a response stream fails, the wait before retrying is now capped to the time remaining, and stalled requests are retried more readily

Unclear The length of the deadline, the minimum time left, and one further condition on retrying are not known.

Details 0 0 Feedback
↑Improved
You'll notice
Useful2 Signal1
MCP unclear

Waiting for MCP servers added mid-session is timed differently#

Claude Code now times its wait for MCP servers added during a session to the moment they answer, and gains a relay hold option

Unclear What sets relayHold and what decides which of the two waiting paths is used are not known.

Details 0 0 Feedback
↑Improved
You'll notice
Useful2 Signal1
Workflows unclear

Workflow agents report a lost reply when the response goes silent#

When a workflow agent's response stream goes silent, Claude Code now reports a lost reply, retries, and gives a specific error if all tries fail

Unclear What makes Claude Code give up on the stream, and the value of the new cap, are not known.

Details 0 0 Feedback
↑Improved
You'll notice
Useful2 Signal1
Background Tasks unclear

Background sessions that fail to start now say what blocked them#

When a background session is blocked from starting, Claude Code now records whether a disclaimer or the folder trust check stopped it

Unclear It is not clear whether the symlink change to the trust check makes any difference you would see.

Details 0 0 Feedback
claude --bg
↑Improved
You'll notice
Useful2 Signal1
SendMessage unclear

Clearer message when cross-machine file transfer is not allowed#

Claude Code now explains that cross-machine file transfer is off because your provider or privacy setup blocks uploads through Anthropic

Unclear It is not clear exactly which provider or privacy settings trigger the message.

Details 0 0 Feedback
↑Improved
You'll notice
Useful2 Signal1
SendMessage unclear

Clearer message when messaging sessions on other machines is blocked#

When sending messages to sessions on another machine is blocked, Claude Code now gives the reason and says local sessions still work

Unclear What causes cross-machine messaging to be blocked is not stated.

Details 0 0 Feedback
↑Improved
You'll notice
Useful2 Signal1
Tools unclear

Error lines shown only for the most recent tool results#

In a group of tool calls, error lines now appear only from the latest real tool result onward instead of for every call

Unclear It is unclear exactly which error lines are now hidden.

Details 0 0 Feedback
↑Improved
You'll notice
Useful2 Signal1
Elsewhere unclear

Away summaries are skipped less often for a stale cache#

The summary shown when you return after being away now uses a second timestamp to judge cache age, so fewer are skipped

Unclear It is not known exactly what the second timestamp records.

Details 0 0 Feedback
↑Improved
You'll notice
Useful1 Signal1
MCP unclear

MCP log message about saving server checks reworded#

A log message about failing to save remembered MCP server checks now uses a shorter prefix and describes a different retry

Unclear Whether the behaviour changed along with the message is not known.

Details 0 0 Feedback
↑Improved
You'll notice
Useful1 Signal1
Elsewhere unclear

Transcripts are processed before being attached to feedback#

When you send feedback or share a session, Claude Code now passes the transcripts through a processing step instead of attaching them as-is

Unclear It is unclear what the processing step does to the transcripts, including whether it removes any data.

Details 0 0 Feedback
↑Improved
You'll notice
Useful1 Signal1
Plugin Marketplace unclear

Plugins are not looked up in a marketplace with a reserved name#

Plugin lookup now finds nothing when the marketplace has a reserved name, and marketplace refusals now carry a reason

Unclear It is unclear which marketplace name is reserved.

Details 0 0 Feedback
↑Improved
You'll notice
Useful1 Signal1
Web Fetch unclear

Suspicious-web-address check gains one more early rejection#

Claude Code's check for suspicious web addresses now rejects one more kind of address before comparing hosts

Unclear It is not clear what kind of address the new early test catches.

Details 0 0 Feedback
↑Improved
You'll notice
Useful1 Signal1
Cloud Sessions

Reworded git and storage messages in cloud sessions#

Cloud session messages now say "screened git" instead of "sync git", and a storage message says "for now" instead of "for this spawn"

Details 0 0 Feedback
↑Improved
You'll notice
Useful1 Signal1
Elsewhere unclear

Conversation compaction handles a switch to a backup model differently#

When compaction hits a switch to a backup model, it now raises that as an error and can put the current model back into the backup list

Unclear It is not clear what condition decides whether the current model is put back into the backup list.

Details 0 0 Feedback
↑Improved
You'll notice
Useful1 Signal1
Permissions unclear

Permission prompts can show a short dimmed note on the title line#

The title row of a permission prompt can now show a dimmed note at its right end when there is room for it

Unclear It is not clear which permission prompts supply this note or what it says.

Details 0 0 Feedback
↑Improved
You'll notice
Useful1 Signal1
Plugin Marketplace unclear

Plugin marketplaces now skip one reserved name#

Claude Code now ignores plugin marketplaces with one reserved name and reads marketplace catalogs slightly differently

Unclear It is not clear what the reserved name is or what the extra input to catalog reading does.

Details 0 0 Feedback
↑Improved
You'll notice
Useful1 Signal1
Status Panel unclear

Account information shows the Profile line differently#

The account information panel now builds the Profile line as text with a suffix in some cases, and shows Organization and Email by a new rule

Unclear It is not clear what the added text says or exactly when the Organization and Email lines now appear.

Details 0 0 Feedback
↑Improved
You'll notice
Useful1 Signal1
Elsewhere unclear

Pull request attribution stats skip quietly on an expected error#

When one kind of error stops pull request attribution stats, Claude Code logs a skip message instead of reporting it as an error

Unclear It is not clear which kind of error leads to the quiet skip.

Details 0 0 Feedback
↑Improved
You'll notice
Useful1 Signal1
Permissions unclear

Shell commands with an unreadable line now ask for approval#

Shell commands with a line Claude Code cannot read now need your approval, and downloaded plugin marketplaces have their name checked

Unclear It is not clear what makes a shell line unreadable, or what happens when a marketplace's name does not match.

Details 0 0 Feedback
↑Improved
You'll notice
Useful1 Signal1
Status Panel unclear

Organizations with HIPAA settings see an extra note#

When your organization's configuration includes HIPAA, an extra dimmed line now appears under the Organization configuration row

Unclear What the extra line says.

Details 0 0 Feedback
↑Improved
You'll notice
Useful1 Signal1
Worktrees unclear

Worktree cleanup reports entries it could not check#

When cleaning up a git worktree, Claude Code now reports entries it could not check instead of silently skipping them

Unclear It is not clear where, or whether, this report is shown to the user.

Details 0 0 Feedback
↑Improved
You'll notice
Useful1 Signal1
Secret Scrubbing unclear

Secret-hiding patterns extended and memory reads recorded differently#

The patterns Claude Code uses to hide secrets were extended, and reads of auto-managed memory are now recorded in a different place

Unclear It is not known what extra text the extended patterns now match.

Details 0 0 Feedback
↑Improved
You'll notice
Useful1 Signal0
Permissions unclear

Small layout changes to skill permission prompt and a confirm dialog#

The skill permission prompt now has padding, one small option row became a confirm-or-cancel dialog, and fork messages show a plain instruction

Unclear It is not clear where in Claude Code the confirm dialog and the fork message are shown.

Details 0 0 Feedback
25 entries

Bug Fixesopen

✓Fixed
You'll notice
Useful3 Signal3
Group of 2 Auth unclear

Claude Code rechecks an unusable sign-in token, plus other sign-in changes#

When the stored sign-in token is unusable, Claude Code rechecks the stored credentials, rate-limited, alongside other sign-in changes

Unclear An extra condition controls when the recheck runs, and it is not clear what it checks or how often the recheck helps in practice.

Details 0 0 Feedback
✓Fixed
You'll notice
Useful3 Signal1
Hooks unclear

Prompts rewritten by a hook now reach Claude as rewritten#

When a hook rewrites your prompt on submit, Claude Code now applies the new text to your message so Claude sees the rewritten version

Unclear It is not certain the old behaviour was actually broken in practice.

Details 0 0 Feedback
✓Fixed
You'll notice
Useful3 Signal1
Sessions unclear

Resumed sessions restart interrupted turns less often by mistake#

On resume, Claude Code checks unanswered tool calls per response and skips restarting a turn whose calls were answered in a restored card

Unclear The exact conditions under which a turn is treated as finished are not clear.

Details 0 0 Feedback
✓Fixed
You'll notice
Useful2 Signal1
Message Queue unclear

Queued messages from different plugins are no longer merged together#

Two queued plugin messages now count as the same only when they come from the same plugin name and the same sender setting

Unclear It is not clear where this comparison runs or whether plugin messages reach it yet.

Details 0 0 Feedback
✓Fixed
You'll notice
Useful2 Signal1
Message Queue unclear

Queued messages are no longer held up by another agent's messages#

Whether a queued message can be delivered now depends only on that agent's own messages, not on other agents' waiting ones

Unclear What the earlier check matched is not known, so it is unclear how often messages were actually held up before.

Details 0 0 Feedback
✓Fixed
You'll notice
Useful1 Signal1
Plugin Marketplace unclear

Built-in plugin marketplace name now matched regardless of capitals#

When scanning plugin marketplaces listed in settings, Claude Code now skips the built-in marketplace's name whatever its capitalisation

Unclear It is not clear what difference the saved-registry detail makes for users.

Details 0 0 Feedback
✓Fixed
You'll notice
Useful1 Signal1
Elsewhere unclear

Bridge heartbeat timing uses its intended interval when at capacity#

When a multisession bridge is at capacity, the next heartbeat is now timed by the heartbeat interval and capped by the at-capacity poll interval

Unclear It is unclear whether this corrects a mistake or deliberately changes the timing.

Details 0 0 Feedback
✓Fixed
You'll notice
Useful1 Signal1
Input unclear

Blank edited prompts no longer send an empty text part#

When an edited prompt's text is blank, Claude Code now keeps only its non-text parts instead of sending an empty piece of text

Unclear It is not shown which feature edits the prompt or what the fallback is.

Details 0 0 Feedback
✓Fixed
You'll notice
Useful1 Signal1
Tools unclear

Unusual tool results are turned into text before they are sent#

Tool results that are neither text nor a list are now converted to text, and cut short if long, before Claude Code sends them

Unclear It is not clear when this conversion runs or whether any setting controls it.

Details 0 0 Feedback
✓Fixed
Under the hood
Useful1 Signal1
Remote Control unclear

Fixes to sign-in header and retry checks when fetching internal events#

A background fetch of internal events now checks the real sign-in headers and uses the correct values for conflict replies and retry waits

Unclear The old code may only have looked wrong after renaming by the build tools, in which case nothing changes in practice.

Details 0 0 Feedback
✓Fixed
You'll notice
Useful1 Signal1
Fullscreen UI unclear

Repeated mouse clicks now reset the double-click count#

In the terminal interface, a repeated mouse click now resets the click count, as a stray click already did

Unclear It is unclear exactly which clicks are marked as repeats.

Details 0 0 Feedback
10 entries

In Developmentopen

9 more of these are in What probably matters to you, on page 1.

◌In dev
Nothing to try yet
Useful3 Signal3
Group of 2 Permissions unclear

Permission dialogs are counted and can show an "N of M" position#

Pending permission prompts, including sandbox network asks, are now counted so each can show its place, such as "1 of 3"

Unclear Where the position label is shown, and whether it is switched on, is not known.

Details 0 0 Feedback
◌In dev
Nothing to try yet
Useful3 Signal3
Auto Mode unclear

Notice explaining why auto mode is off, built but off by default#

Claude Code can now work out why auto mode is off and build a notice saying so, behind a remote setting that falls back to off

Unclear It is not clear whether the remote setting controls the notice, or where the notice appears.

Details 0 0 Feedback
◌In dev
Nothing to try yet
Useful3 Signal3
Background Tasks unclear

Internal support for restoring a saved task list#

New code can restore a task list in one step, with an option to roll back entirely if any task fails

Unclear Nothing shows what calls this code or when it runs.

Details 0 0 Feedback

Continues on page 3 →

Verbatim
Official · Anthropic

Anthropic’s official release notes

Published verbatim by Anthropic for v2.1.286. 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 88 bullets, 14 name something an entry on this page also names, 16 name something no entry here does, and 58 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.

  • Added a count such as "2 of 5" to the permission prompt when several permission requests stack up No entry names this
  • Added mouse support for the "N more" rows of lists in fullscreen mode: click one to jump to that end of the list, with hover and pressed states Probably non-string-tool-results-get-converted-to-text-before-sending
  • Fixed several Claude Code processes and IDE extensions each opening a login browser when gcpAuthRefresh or awsAuthRefresh credentials expire Nothing to match on
  • Fixed claude --resume and --continue sometimes losing every turn after a batch of parallel tool calls when the earlier session crashed or was killed No entry names this
  • Fixed API 400 errors after a tool or hook returned an object, number or boolean instead of text, including in resumed sessions Nothing to match on
  • Fixed cloud sessions with very large histories never waking up because the container was stopped while the transcript was still loading Nothing to match on
  • Fixed the Claude apps gateway's spend meter pricing 1-hour prompt cache writes at the cheaper 5-minute rate, and counting only the first model call's input tokens on streamed turns that run a server-side tool such as web search Nothing to match on
  • Fixed macOS sessions still showing "Not logged in" or "Login expired" after /login succeeds in another Claude Code window when a leftover ~/.claude/.credentials.json exists Probably auth-source-classification-changed-login-managed-key, ultrareview-stop-message-and-other-small-changes, oauth-provider-docs-tls-and-client-guarantees-text-updated
  • Fixed every turn failing when the Anthropic API refuses the model your default or a model alias resolves to: Claude Code now retries once on the previous model of the same tier Nothing to match on
  • Fixed Remote Control sessions (including claude remote-control) staying connected after your organization's policy turns Remote Control off; they now disconnect with a notice Probably remote-control-tears-down-when-org-policy-starts-refusing-it
  • Fixed refusal and --fallback-model retries failing when the fallback model can't run fast; they now run at standard speed, with a one-time notice in interactive sessions No entry names this
  • Fixed headless sessions repeating the "MCP servers require authentication" reminder after a successful re-authentication when the MCP discovery cache is enabled No entry names this
  • Fixed claude auth status reporting a Console sign-in's stored API key as claude.ai; it now reports api_key, and the VS Code extension treats that session as an API key session Probably auth-source-classification-changed-login-managed-key, mcp-discover-prior-tracking-generalized-beyond-claudeai-prox
  • Fixed /status listing an Anthropic profile beside an API key as if both were in effect; the profile is now marked as not in use No entry names this
  • Fixed Claude not being told when a file attached to a message sent over Remote Control did not arrive, and a file sometimes getting only 10 seconds for its last download try Nothing to match on
  • Fixed a Remote Control message that arrived while Claude Code was exiting being marked delivered and then never answered; it now stays queued for the session's next run Nothing to match on
  • Fixed MCP error messages showing a credential's value when "Bearer" or "Basic" came before its key name Probably bearer-token-redaction-regex-handles-percent-encoded-tokens
  • Fixed percent-encoded Bearer tokens being only partly masked in error messages Nothing to match on
  • Fixed redacted logs and transcripts showing a secret whose key name has an invisible character inside, such as a zero-width space Nothing to match on
  • Fixed logs and transcripts showing part of a URL password that contains punctuation such as ), quotes, ], & or a second @, or that runs past a / to a bracketed host such as [::1] in an ssh URL No entry names this
  • Fixed the session transcript in the zip that /feedback saves to disk containing invalid JSON lines after secret redaction No entry names this
  • Fixed MCP connectors listing no tools for up to a day after their server dropped the older MCP handshake Nothing to match on
  • Fixed a repeat MCP sign-in request from Claude replacing the pending sign-in link, which could stop that link from working Nothing to match on
  • Fixed /usage not crediting an MCP server for a tool call made while that server was still connecting or had only just connected, such as right after a restart No entry names this
  • Fixed plugins enabled on claude.ai occasionally going missing from Claude Code for a session after a transient server error Nothing to match on
  • Fixed a message typed into a running subagent showing twice in its transcript after the subagent read it Nothing to match on
  • Fixed subagent hand-back messages showing a raw task id instead of the agent's name when the subagent had no registered name Nothing to match on
  • Fixed foreground subagents sometimes missing the task-tracking tools (TaskCreate/Get/Update/List, TodoWrite) in sessions that have them enabled Nothing to match on
  • Fixed subagents spawned with worktree isolation loading the project CLAUDE.md and its imports a second time from the worktree copy on their first file read Nothing to match on #98052Worktree-isolated subagents load the project CLAUDE.md (and its @imports) twice; the only workaround exclude pattern that doesn't drop subfolder CLAUDE.md files is non-obvious Closed
  • Fixed Workflow tool subagents being restarted from their original prompt when a connection stalled for a few minutes mid-response Nothing to match on
  • Fixed /compact, /clear, and /rewind typed while viewing a background agent's or teammate's transcript silently acting on the main conversation: a dialog now names the target and asks first Probably agent-view-slash-commands-compact-clear-rewind-now-prom, fallback-model-switch-warns-about-a-smaller-context-window #97857[BUG] `/compact` typed inside the subagent view compacts the main conversation instead Open
  • Fixed background jobs showing done while waiting for your approval Nothing to match on
  • Fixed the commit attribution reminder being re-sent inside tool output when a model fallback lasts only one turn Nothing to match on #92386Unrelated <system-reminder> injected into Bash tool result, instructing to add Claude attribution to commits Open
  • Fixed a click on the space between words of a collapsed row (such as "Thought for 4s") highlighting the row without expanding it in fullscreen mode No entry names this
  • Fixed a row with no details, such as an action row with a long name, pushing every other row's details to the right in list screens Nothing to match on
  • Fixed files with very long names not reaching a cloud session when attached to it Nothing to match on
  • Fixed plugin errors for a marketplace Claude Code refuses to load: they now say why and how to fix it instead of "not found" No entry names this
  • Fixed /plugin's Discover tab showing a marketplace name unquoted in its "Checking … for new plugins" line when its rows already show that name in quotes Probably hook-descriptions-reworded, plugin-installenabledisableuninstall-in-plugin-ui-now-re, claude-test-plugin-conflict-message-and-runner-rewritten, plugin-types-command-hidden, you-should-know-builtin-plugin-side-agent-suggests-things
  • Improved commit guidance: when your project or user skills include one named verify, Claude is now told to run it right before committing, except for docs-only and tests-only commits Nothing to match on
  • Improved send now (ctrl+enter) in a subagent's view: it now moves the subagent's running command to the background so your message is read right away Nothing to match on
  • Improved replies from background agents to your messages so they no longer open with a separate recap of what you said Nothing to match on
  • Improved claude.ai artifact link reads: WebFetch now asks the same questions as the Artifact tool's read (no artifact prompt while the session's network access is on, one per artifact while it is off), and an auto-mode yes no longer counts where only you can answer Nothing to match on
  • Improved fetch, skill, file read, sandbox network, Claude in Chrome, workflow script and notebook edit permission prompts to match the look of file edit prompts Nothing to match on
  • Improved Bash, PowerShell and Monitor permission prompts to show the command between dashed lines, matching file edit prompts Nothing to match on
  • Improved list scrollbars in fullscreen mode: in most lists the bar no longer shifts as the "N more" rows come and go, and it now has ↑/↓ arrows you can click or hold to scroll Probably non-string-tool-results-get-converted-to-text-before-sending
  • Improved the external editor (Ctrl+G): editors that take a line number now open on the line your cursor is on in the prompt Nothing to match on
  • Improved slash command suggestion responsiveness while typing when many skills or plugin commands are installed; command descriptions now match by word prefix Nothing to match on
  • Improved the output style picker: it now opens on your current style instead of Default, with each style's description on the line under its name; number keys no longer pick a style Nothing to match on
  • Improved /hooks: the closing line of a hook's detail screen now says "this hook" instead of "it" Probably hooks-dialog-now-opens-on-a-flat-list-of-configured-hooks
  • Improved the model fallback notice and the autocompact-thrashing error to say when a fallback dropped the context window from 1M to 200K tokens Nothing to match on
  • Improved responsiveness of SDK and -p sessions when a host re-sends an MCP server enable for a server that is already connected Nothing to match on
  • Improved the protocol page a Claude apps gateway serves at /protocol: it now says not to reject unknown input and matches what Claude Code sends today No entry names this
  • Changed prompts sent while nothing is running or queued to show in the normal text color right away instead of gray Nothing to match on
  • Changed how failed API requests are retried: one limit now covers a whole model call, so with the default retry settings a failing call sends at most 14 requests Nothing to match on
  • Changed --bare to connect only the MCP servers named on the command line, send the model no system reminders, and start no background tasks; under --bare, a shell command that reaches its timeout now stops instead of moving to the background Probably bare-sessions-refuse-plugin-toolregister
  • Changed the send-now key (ctrl+enter) to move a skill's own shell command to the background instead of ending it Nothing to match on
  • Changed the WebFetch error for a rate-limited domain safety check to tell Claude not to retry it in a loop Nothing to match on
  • Changed plugin installs to refuse npm sources that are git repositories or folders, and to install plugin dependencies only from registry packages Nothing to match on
  • Changed list screens (/artifacts, /mcp, /skills, /hooks and others) to always line up each row's details in one column after the names Probably mcp-oauth-tool-reuses-an-in-progress-sign-in, mcp-reconnect-and-ccr-diagnostics-additions, hooks-dialog-now-opens-on-a-flat-list-of-configured-hooks
  • Changed the overflow rows of lists to read "↑ N more" / "↓ N more" instead of "N more above" / "N more below" No entry names this
  • Changed /hooks to open on one list of your configured hooks grouped by event, so viewing a hook takes one Enter instead of three Probably hooks-dialog-now-opens-on-a-flat-list-of-configured-hooks
  • Changed the theme picker to a scrolling list that fits your terminal instead of pushing the preview off screen; number keys no longer pick a theme Nothing to match on
  • Changed /exit's Remove worktree to run after Claude Code stops the servers and shells it started there, which on Windows could keep the folder from being deleted No entry names this
  • Changed the claude-api skill's Managed Agents examples to create environments with limited networking No entry names this
  • Removed the browser link from /ultrareview and claude ultrareview output Probably ultrareview-stop-message-and-other-small-changes, stop-ultrareview-dialog-uses-confirmcancel-control-with-key, ultrareview-detail-dialog-drops-open-in-browser-and-shows, ultrareview-launch-message-can-say-findings-arrive-in-the-se
  • Windows: Fixed claude --bg and the agents view refusing a folder that claude already trusts when its trust record was saved with different letter case Probably background-session-workspace-trust-and-gate-blocked-reasons
  • [VSCode] Added bookmarks: save Claude's responses and keep them in view in a Bookmarks side panel Nothing to match on
  • [VSCode] Added the questions Claude asks and your answers to the conversation: after you answer a question card, a Questions row shows each question with your picks Nothing to match on
  • [VSCode] Added option previews to question cards in the chat panel: the highlighted choice's mockup or snippet shows beside or under the options Nothing to match on
  • [VSCode] Added rows under a message that open to the terminal output, browser tab, browser instructions and selected code sent with it Nothing to match on
  • [VSCode] Fixed a second copy of a conversation opening in a tab when it was already open in the side bar; the side bar now switches to it Nothing to match on
  • [VSCode] Fixed settings dialogs reporting a failed save, without re-checking, when Claude Code printed more than 1 MB of output Nothing to match on
  • [VSCode] Fixed an endless "Teleporting session…" spinner when the extension stops responding No entry names this
  • [VSCode] Improved the Manage plugins dialog: it says when a turned-off plugin is still on because of other settings, and explains a plugin folder clash Nothing to match on
  • [VSCode] Changed Stop and Escape to end only the current turn; background agents keep running and can be stopped one by one from the agent map Nothing to match on
  • [VSCode] Changed the "✻ Claude Code" status bar item to show in every window, so you can open Claude when no file is open No entry names this
  • [Cloud sessions] Fixed an answered question card or approved tool call getting no reply when the session had gone idle after Claude sent a message Nothing to match on
  • [Cloud sessions] Fixed clearing an organization environment's setup script in admin settings leaving new cloud sessions still running the old script Nothing to match on
  • [Cloud sessions] Fixed the Runner actions menu on the self-hosted environments admin page closing on its own a few seconds after it opened Nothing to match on
  • [Cloud sessions] Fixed routine runs whose cloud session never started showing as Succeeded in the Runs pane, the routine's page and the sidebar; they now show as Failed Nothing to match on
  • [Cloud sessions] Fixed clicking an audio or video file in a cloud session's Outputs card opening an empty file search instead of playing the file Nothing to match on
  • [Cloud sessions] Changed a routine's page to read "Due" with the scheduled time, instead of a next run time in the past, when a scheduled run is late and hasn't started Nothing to match on
  • [Claude Tag] Added an Add channel button to Claude Tag's spend limits page in admin settings, so a limit can be set on any channel, including a private one, from its channel ID or Slack link Nothing to match on
  • [Claude Tag] Fixed memory recall finding nothing in organizations that cannot use the default Sonnet model Nothing to match on
  • [Claude Tag] Fixed a Slack channel losing its Claude settings when an Enterprise Grid admin moves it to another workspace and the first post afterward doesn't mention Claude Nothing to match on
  • [Claude Tag] Fixed Claude in a Slack thread occasionally starting over on a fresh machine, losing work it hadn't pushed, when your reply answered a question it had just asked Nothing to match on
  • [Claude Tag] Fixed public channel names on Claude Tag's spend limits page in admin settings showing as raw Slack IDs in larger organizations Nothing to match on
  • [Claude Tag] Improved the titles of sessions started from Slack as shown on claude.ai: they now read as the words you typed, without Slack user IDs or escape codes Nothing to match on

A model matched these bullets to the GitHub issues they fix, so a link can be wrong.

System prompt

The appended system-reminder blocks moved: 1 line added.

Claude Code, interactive mode

Documentation

What the docs did around this release

928 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.

Switches

Every name in this release

The 66 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.

Slash commands

CLI flags

Environment variables

Settings and names in the code

Feedback