Anthropic, Claude Code changelog

Worauf sich diese Seite beruft, wörtlich, abgerufen am 12.10.2026:

• Agents and bash commands can run asynchronously and send messages to wake up the main agentbestätigt 10.11.2026
• pressing Ctrl+B now backgrounds all running foreground tasks simultaneouslybestätigt 10.11.2026
• Subagents now run in the background by default, so Claude keeps working while they run and is notified when they finishbestätigt 10.11.2026
• a single list of every Claude Code sessionbestätigt 10.11.2026
• Background agents now continue running when you press ESC to cancel the main threadbestätigt 10.11.2026
• approvals granted in a git worktree persist across sessions and worktreesbestätigt 10.11.2026
• allowing agents to declaratively run in isolated git worktreesbestätigt 10.11.2026
• now stay alive when idle, are restarted in place to apply Claude Code updatesbestätigt 10.11.2026
• now appear alongside interactive ones, marked with bgbestätigt 10.11.2026
• Added /loop command to run a prompt or slash command on a recurring intervalbestätigt 10.11.2026
• Added cron scheduling tools for recurring prompts within a sessionbestätigt 10.11.2026
• Added Monitor tool for streaming events from background scriptsbestätigt 10.11.2026
• Added the Monitor tool and plugin monitors on Bedrock, Vertex, and Foundry, and when telemetry or nonessential traffic is disabled; long foreground sleeps are now blocked there toobestätigt 10.11.2026
• set a completion condition and Claude keeps working across turns until it's metbestätigt 10.11.2026
• Send background tasks to Claude Code on the web by starting a message with &bestätigt 10.11.2026
• Expand Remote Control to more usersbestätigt 10.11.2026
• point it at a marker file to suppress mobile push notifications while you're at the machinebestätigt 10.11.2026
• Claude Code now continues your session automatically when a claude.ai usage limit resetsbestätigt 10.11.2026
• when background tasks keep a goal waiting for 30+ minutes, Claude now checks in on them instead of waiting indefinitelybestätigt 10.11.2026
• Added background monitor support for plugins via a top-level monitors manifest keybestätigt 10.11.2026
• Reverted the 2.1.290 fix for cloud sessions staying asleep after a container restart lost a pending /loop wakeup or scheduled taskbestätigt 10.11.2026
• so runaway or chatty /loop tasks are easy to spotbestätigt 10.11.2026
• to report more or fewer findings than the usual limitbestätigt 10.11.2026
• The old cleanup-and-fix behavior has been removedbestätigt 10.11.2026
• now runs a cleanup-only review (reuse, simplification, efficiency, altitude) and applies the fixesbestätigt 10.11.2026
• for a deep cloud reviewbestätigt 10.11.2026
• comprehensive code review in the cloud using parallel multi-agent analysis and critiquebestätigt 10.11.2026
• merged five cleanup finders into one, cutting token usage by roughly 25%bestätigt 10.11.2026
• Changed /code-review to use leaner inline review prompts for every model that has no tuned settings of its own, instead of spawning many review subagentsbestätigt 10.11.2026
• New native VS Code extensionbestätigt 10.11.2026
• Added full markdown document view for plans in VS Code, with support for adding comments to provide feedbackbestätigt 10.11.2026
• Introducing Claude Code for Desktopbestätigt 10.11.2026
• Claude Code Web: Support for Web -> CLI teleportbestätigt 10.11.2026
• Added /teleport and /remote-env slash commands for claude.ai subscribers, allowing them to resume and configure remote sessionsbestätigt 10.11.2026
• Added claude --desktop to open the Claude desktop app on the current directorybestätigt 10.11.2026
• subcommand for external builds, enabling local environment serving for all usersbestätigt 10.11.2026
• Added prompt-based stop hooksbestätigt 10.11.2026
• Added /team-onboarding command to generate a teammate ramp-up guide from your local Claude Code usagebestätigt 10.11.2026
• Added recap feature to provide context when returning to a sessionbestätigt 10.11.2026
• Claude Code sessions can now message each other, on any of your machines, with ListAgents to discover thembestätigt 10.11.2026
• Hardened cross-session messaging: messages relayed via SendMessage from other Claude sessions no longer carry user authoritybestätigt 10.11.2026
• messages sent to other agent sessions via SendMessage are now evaluated by the permission classifier before dispatchbestätigt 10.11.2026
• Windows: cross-session messaging is now availablebestätigt 10.11.2026
• Added attribution setting to customize commit and PR bylinesbestätigt 10.11.2026
• Added recognition of glab mr create/merge/close/reopen/note/updatebestätigt 10.11.2026
• Added session URL attribution to commits and PRs created from web sessionsbestätigt 10.11.2026
• Added attribution.sessionUrl setting to omit the claude.ai session link from commits and PRs in web and Remote Control sessionsbestätigt 10.11.2026
• the glab CLI config store gets the same sandbox and credential-path protection as ghbestätigt 10.11.2026
• --from-pr now accepts GitLab merge-request, Bitbucket pull-request, and GitHub Enterprise PR URLsbestätigt 10.11.2026
• Pasting a PR URL into the /resume search box now finds the session that created that PR (GitHub, GitHub Enterprise, GitLab, and Bitbucket)bestätigt 10.11.2026
• Added prUrlTemplate setting to point the footer PR badge at a custom code-review URL instead of github.combestätigt 10.11.2026
• Added a GitLab merge request badge to the footer and statuslinebestätigt 10.11.2026
• GitHub Actions workflow setup is now optionalbestätigt 10.11.2026
• Added a warning in /web-setup when the GitHub CLI token lacks the workflow scopebestätigt 10.11.2026
• Added GitLab support to plugin marketplacesbestätigt 10.11.2026
• Added GitLab merge request URL support to the --worktree flagbestätigt 10.11.2026
• send the model no system reminders, and start no background tasksbestätigt 10.11.2026
• Released TypeScript SDK: import @anthropic-ai/claude-code to get startedbestätigt 10.11.2026
• SDK: The Claude Code SDK is now the Claude Agent SDKbestätigt 10.11.2026
• turns your own machines or containers into a place Claude Code web, mobile, and desktop sessions can run, on Team and Enterprise plansbestätigt 10.11.2026
• environment variable to include subagent text and thinking in stream-json outputbestätigt 10.11.2026
• for unattended headless hosts: anything that would prompt is denied automatically while the active permission mode (including auto mode) keeps decidingbestätigt 10.11.2026
• Shared project permission rules can be saved in .claude/settings.jsonbestätigt 10.11.2026
• syntax to set any setting from the promptbestätigt 10.11.2026
• Settings file changes take effect immediatelybestätigt 10.11.2026
• to start Claude Code with all customizationsbestätigt 10.11.2026
• to limit which API providers a machine may usebestätigt 10.11.2026
• can no longer turn it on (they can still turn it off)bestätigt 10.11.2026
• setting to configure Claude’s response languagebestätigt 10.11.2026
• letting separate teams deploy independent policy fragments that merge alphabeticallybestätigt 10.11.2026
• present but not applied because a higher-precedence managed source is activebestätigt 10.11.2026
• Add support for enterprise managed MCP allowlist and denylistbestätigt 10.11.2026
• Added server-managed settings diagnosticsbestätigt 10.11.2026
• Deprecated claude config commands in favor of editing settings.jsonbestätigt 10.11.2026
• Unshipped claude config commandbestätigt 10.11.2026
• Managed settings can now be set via macOS plist or Windows Registry.bestätigt 10.11.2026
• enabling enterprise security auditing and optional blocking of settings changesbestätigt 10.11.2026
• Claude Code refuses to start if its version is outside the allowed rangebestätigt 10.11.2026
• works in interactivebestätigt 10.11.2026
• Add support for enterprise managed settings. Contact your Anthropic account team to enable this feature.bestätigt 10.11.2026
• Removed deprecated Windows managed settings fallback atbestätigt 10.11.2026
• scope now allows you to add MCP servers to .mcp.json files and commit them to your repositorybestätigt 10.11.2026
• Remote MCP servers (SSE and HTTP) now support OAuthbestätigt 10.11.2026
• Added claude mcp login <name> and claude mcp logout <name> to authenticate MCP servers from the CLI without opening the interactive /mcp menubestätigt 10.11.2026
• Added support for using claude.ai MCP connectors in Claude Codebestätigt 10.11.2026
• Added pre-configured OAuth client credentials for MCP servers that don’t support Dynamic Client Registrationbestätigt 10.11.2026
• runs without inherited credential env varsbestätigt 10.11.2026
• Changed the default limit on MCP tool descriptions sent up front and on MCP server instructions from 2,048 to 4,096 charactersbestätigt 10.11.2026
• Changed MCP tool descriptions the model loads through tool search to be cut at 16,384 characters instead of 2,048bestätigt 10.11.2026
• MCP servers can now request structured input mid-task via an interactive dialog (form fields or browser URL)bestätigt 10.11.2026
• Added syncing of the skills and plugins enabled on your claude.ai account to terminal sessions signed in with itbestätigt 10.11.2026
• MCP tool descriptions and server instructions are now capped at 2KB to prevent OpenAPI-generated servers from bloating contextbestätigt 10.11.2026
• Added CLAUDE_CODE_MAX_MCP_DESCRIPTION_LENGTH to change the 2,048-character cap on MCP tool descriptions and server instructions for every MCP server in the sessionbestätigt 10.11.2026
• Changed plugin marketplaces whose name imitates a reserved marketplace name to be refused when addedbestätigt 10.11.2026
• so a package’s install scripts no longer runbestätigt 10.11.2026
• Added native plugin management supportbestätigt 10.11.2026
• Changed /recap to decline with a short notice when it arrives relayed from a chat thread (your own included) or from a routine or webhookbestätigt 10.11.2026
• typed in the terminal, the Claude apps, Remote Control, -p or an SDK host, it runs as beforebestätigt 10.11.2026
• use SendMessage({to: agentId}) to continue a previously spawned agentbestätigt 10.11.2026
• Added support for .claude/rules/bestätigt 10.11.2026
• Added AGENTS.md support: in a project with no CLAUDE.md, Claude Code reads AGENTS.md insteadbestätigt 10.11.2026
• Claude now automatically records and recalls memories as it worksbestätigt 10.11.2026
• CLAUDE.md files can now import other files.bestätigt 10.11.2026
• Improved the external CLAUDE.md imports approval dialog to show which files are being imported and from wherebestätigt 10.11.2026
• Added support for loading CLAUDE.md files from additional directories specified via --add-dir flagbestätigt 10.11.2026
• Changed AGENTS.md support to also work on Amazon Bedrock, Google Vertex AI, Microsoft Foundry, LLM gateways, and sessions with telemetry disabledbestätigt 10.11.2026
• no longer triggers the security approval dialogbestätigt 10.11.2026
• Changed /memory so that Auto-memory can no longer be turned on from a background sessionbestätigt 10.11.2026
• Added autoMemoryDirectory setting to configure a custom directory for auto-memory storagebestätigt 10.11.2026
• Added Memory and Instructions entries to the Customize menubestätigt 10.11.2026
• index now truncates at 25KB as well as 200 linesbestätigt 10.11.2026
• the agent is now reminded to compact its MEMORY.md index when nearing the size limitbestätigt 10.11.2026
• Memory writes that leave a MEMORY.md index over its read limit now produce an explicit errorbestätigt 10.11.2026
• warning threshold now scales with the model’s context windowbestätigt 10.11.2026
• letting custom and plugin subagents run without user, project and local CLAUDE.md files; managed policy files still loadbestätigt 10.11.2026
• /memory command now allows direct editing of all imported memory filesbestätigt 10.11.2026
• Added memory frontmatter field support for agentsbestätigt 10.11.2026
• Claude automatically saves useful context to auto-memory. Manage with /memorybestätigt 10.11.2026
• Project configs & auto memory now shared across git worktrees of the same repositorybestätigt 10.11.2026
• Added InstructionsLoaded hook event that fires when CLAUDE.md or .claude/rules/*.md files are loaded into contextbestätigt 10.11.2026
• Added last-modified timestamps to memory filesbestätigt 10.11.2026
• Added an ISO modified timestamp to memory file frontmatterbestätigt 10.11.2026
• Added /doctor prompt-auditbestätigt 10.11.2026
• Quickly add to Memory by starting your message withbestätigt 10.11.2026
• Fixed issue where pasting could trigger memory or bash mode unexpectedlybestätigt 10.11.2026
• Removed # shortcut for quick memory entrybestätigt 10.11.2026
• to be hidden from Claude when auto-injectedbestätigt 10.11.2026
• Markdown files in .claude/commands/ directories now appear as custom slash commands to insert prompts into your conversationbestätigt 10.11.2026
• Merged slash commands and skills, simplifying the mental model with no change in behaviorbestätigt 10.11.2026
• Added SlashCommand tool, which enables Claude to invoke your slash commands.bestätigt 10.11.2026
• The model can now discover and invoke built-in slash commands likebestätigt 10.11.2026
• Custom slash commands: Run bash output, @-mention files, enable thinking with thinking keywordsbestätigt 10.11.2026
• setting to disable inline shell execution in skills, custom slash commands, and plugin commandsbestätigt 10.11.2026
• Added support for running skills and slash commands in a forked sub-agent context usingbestätigt 10.11.2026
• to run in the background by default; opt out per skill withbestätigt 10.11.2026
• frontmatter support for skills and slash commands to override the model effort level when invokedbestätigt 10.11.2026
• Changed skills without additional permissions or hooks to be allowed without requiring approvalbestätigt 10.11.2026
• environment variable to hide bundled skills, workflows, and built-in slash commands from the modelbestätigt 10.11.2026
• skills created or modified in ~/.claude/skills or .claude/skills are now immediately available without restarting the sessionbestätigt 10.11.2026
• command to re-scan skill directories without restarting the sessionbestätigt 10.11.2026
• on your machine their bodies don’t run ! commands or expand @ filesbestätigt 10.11.2026
• skill for building applications with the Claude API and Anthropic SDKbestätigt 10.11.2026
• Reduced the context cost of loading the built-in claude-api skill from ~200k+ tokens to ~25k by loading reference docs on demandbestätigt 10.11.2026
• Skill character budget now scales with context window (2% of context)bestätigt 10.11.2026
• listing are now capped at 250 characters to reduce context usagebestätigt 10.11.2026
• Claude no longer runs the /verify and /code-review skills on its ownbestätigt 10.11.2026
• Added a tip on Bash permission prompts pointing to auto modebestätigt 10.11.2026
• Added wildcard pattern matching for Bash tool permissions using * at any position in rulesbestätigt 10.11.2026
• Ask permissions: have Claude Code always ask for confirmation to use specific tools with /permissionsbestätigt 10.11.2026
• Auto mode is now available for Max subscribers when using Opus 4.7bestätigt 10.11.2026
• Auto mode no longer requires --enable-auto-modebestätigt 10.11.2026
• Auto mode no longer requires opt-in consentbestätigt 10.11.2026
• interactive terminal and VS Code sessions to start in auto mode when no permission mode is configured, on every plan and providerbestätigt 10.11.2026
• Releasing a sandbox mode for the BashTool on Linux & Macbestätigt 10.11.2026
• credentials setting to block sandboxed commands from reading credential files and secret environment variablesbestätigt 10.11.2026
• since they also match options inserted before the subcommandbestätigt 10.11.2026
• Added Tool(param:value) syntax for permission rules to match a tool’s input parametersbestätigt 10.11.2026
• Added wildcard syntax mcp__server__* for MCP tool permissions to allow or deny all tools from a serverbestätigt 10.11.2026
• denied commands now show a notification and appear in /permissionsbestätigt 10.11.2026
• Changed auto mode to no longer read autoMode from .claude/settings.local.jsonbestätigt 10.11.2026
• to add custom rules alongside the built-in list instead of replacing itbestätigt 10.11.2026
• for auto mode classifier rules that block unconditionally regardless of user intent or allow exceptionsbestätigt 10.11.2026
• Added claude auto-mode reset to restore the default auto-mode configurationbestätigt 10.11.2026
• Added auto-mode denial reasons to the transcriptbestätigt 10.11.2026
• to default to the server-side classifier, which does not charge for classifier overheadbestätigt 10.11.2026
• Added a prompt before writing to shell startup filesbestätigt 10.11.2026
• which could otherwise lead to unintended command executionbestätigt 10.11.2026
• removes the built-in tools that run commands or code and WebFetchbestätigt 10.11.2026
• New /approved-tools command for managing tool permissionsbestätigt 10.11.2026
• Renamed /allowed-tools -> /permissionsbestätigt 10.11.2026
• are now blocked when you didn’t ask to discard local workbestätigt 10.11.2026
• setting to route all Bash/PowerShell commands through the auto-mode classifierbestätigt 10.11.2026
• Added an Auto mode tab to /permissions for viewing and editing auto mode classifier rulesbestätigt 10.11.2026
• scans transcripts for common read-only Bash and MCP tool calls and proposes a prioritized allowlistbestätigt 10.11.2026
• Released hooks.bestätigt 10.11.2026
• Added Claude Mods: plugins may now modify deeper behaviorbestätigt 10.11.2026
• (for first-party sessions with telemetry on)bestätigt 10.11.2026
• Added conditional if field for hooks using permission rule syntaxbestätigt 10.11.2026
• PreToolUse hooks can now modify tool inputsbestätigt 10.11.2026
• PostToolUse hooks can now replace tool output for all tools viabestätigt 10.11.2026
• Added hooks support for skill and slash command frontmatterbestätigt 10.11.2026
• Added a Hooks dialog to the command menu for viewing hooks and adding, editing, or removing them in user, project, and local settingsbestätigt 10.11.2026
• Special thanks to community input inbestätigt 10.11.2026
• a hook that can’t start, times out, or exits with an unexpected code blocks the action instead of letting it throughbestätigt 10.11.2026
• naming the approval an organization requires for a toolbestätigt 10.11.2026
• that spawns the command directly without a shell, so path placeholders never need quotingbestätigt 10.11.2026
• Improvements to plan modebestätigt 10.11.2026
• Added /plan command shortcut to enable plan mode directly from the promptbestätigt 10.11.2026
• users can now enter plan mode without approvalbestätigt 10.11.2026
• the auto-mode classifier judges them insteadbestätigt 10.11.2026
• setting to customize where plan files are storedbestätigt 10.11.2026
• Added feedback input when rejecting plans, allowing users to tell Claude what to changebestätigt 10.11.2026
• Plan mode: introduced new Plan subagentbestätigt 10.11.2026
• Ask Claude to make a plan with thinking modebestätigt 10.11.2026
• Claude will now ask you questions more often in plan modebestätigt 10.11.2026
• Plan Mode now builds more precise plans and executes more thoroughlybestätigt 10.11.2026
• Sessions are now auto-named from plan content when you accept a planbestätigt 10.11.2026
• Removed ultraplan featurebestätigt 10.11.2026
• Added a /doctor check that proposes trimming checked-inbestätigt 10.11.2026
• Added LSP (Language Server Protocol) tool for code intelligence features like go-to-definition, find references, and hover documentationbestätigt 10.11.2026
• Introducing the Explore subagentbestätigt 10.11.2026
• Bash deny rules now match commands wrapped inbestätigt 10.11.2026
• Skills and slash commands can now set disallowed-tools in frontmatter to remove tools from the model while the skill is activebestätigt 10.11.2026
• explicitly listing Grep/Glob now provides the dedicated search tools on native builds with embedded search (previously these names were silently ignored)bestätigt 10.11.2026
• Released output stylesbestätigt 10.11.2026
• Added support for Claude Skillsbestätigt 10.11.2026
• Added research preview agent teams feature for multi-agent collaborationbestätigt 10.11.2026
• Un-deprecate output styles based on community feedbackbestätigt 10.11.2026
• The lean system prompt is now the default for all models except Haiku, Sonnet, and Opus 4.7 and earlierbestätigt 10.11.2026
• Changed interactive sessions on third-party providers or with telemetry off to start in auto modebestätigt 10.11.2026
• Improved /code-review --comment to post findings on GitLab merge requests via glab mr notebestätigt 10.11.2026
• they now link to the gitlab.com issuebestätigt 10.11.2026
• Fixed /autofix-pr reportingbestätigt 10.11.2026
• breaking headless sessions until restartbestätigt 10.11.2026
• Fixed --json-schema silently producing unstructured output when the schema was invalidbestätigt 10.11.2026
• Fixed an active /goal being lost when resumingbestätigt 10.11.2026
• Fixed scheduled tasks (/loop with an interval, reminders) silently not coming back on resume once the conversation was compacted; covers compactions made from this version onbestätigt 10.11.2026
• Fixed scheduled tasks set in the foreground never firing after abestätigt 10.11.2026
• they now classify as task notifications and can no longer approve a pending actionbestätigt 10.11.2026
• previously only Read loaded thembestätigt 10.11.2026
• Fixed path-scoped rules and nested CLAUDE.md files not loading when Claude views a file with a single-file cat, head, tail, sed -n or grep command in the Bash tool instead of the Read toolbestätigt 10.11.2026
• brace expansion is now budget-boundedbestätigt 10.11.2026
• Improved the large CLAUDE.md startup notice to also count instruction files togetherbestätigt 10.11.2026
• invisible characters and tags that imitate Claude Code’s own markup are neutralized inbestätigt 10.11.2026
• from outside the project being skipped without ever showing the external-imports approval promptbestätigt 10.11.2026
• reviews now use its version from the base branchbestätigt 10.11.2026
• on custom commands and skills being ignored in interactive sessionsbestätigt 10.11.2026
• frontmatter on custom commands, skills, and subagents being ignored on models whose default effort is still pinnedbestätigt 10.11.2026
• raised the listing cap from 250 to 1,536 charactersbestätigt 10.11.2026
• 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 commitsbestätigt 10.11.2026
• a 200K+ token conversation left idle for about 54 minutes is compacted before its one-hour prompt cache expiresbestätigt 10.11.2026
• Changed claude -p and Agent SDK sessions with no permission mode to start in default mode on third-party providers, with telemetry off or on a fresh installbestätigt 10.11.2026
• Changed claude -p and Python Agent SDK sessions on third-party providers or with telemetry off to start in auto mode when no permission mode is configuredbestätigt 10.11.2026
• row to /config for turning idle compaction off or back onbestätigt 10.11.2026
• Fixed scheduled tasks refusing their own configured prompt as untrusted inputbestätigt 10.11.2026
• Changed mods’ settings.read hooks to receive each settings layer and rewrite the merged settings, so a managed mod can make an organization’s setting a ceilingbestätigt 10.11.2026
• including the cleanupPeriodDays the retention sweep usesbestätigt 10.11.2026
• to no longer keep their allowed-tools when your organization’s managed settings set allowManagedPermissionRulesOnlybestätigt 10.11.2026
• Changed UserPromptSubmit hooks to run before a skill or custom slash command expands, so a hook that blocks the prompt now also stops its embedded shell commandsbestätigt 10.11.2026
• updatedInput from PreToolUse and PermissionRequest hooks is now validated like Claude’s own input: unknown keys are dropped, and input the tool rejects denies the callbestätigt 10.11.2026
• when a plugin that runs first skips one of your mod’s hooks, naming the event and the pluginbestätigt 10.11.2026
• Changed a mod hook’s .catch handler to run only when the hook throws or times outbestätigt 10.11.2026
• when the hook cannot be handed them or cannot hand back its answerbestätigt 10.11.2026
• a tool that already ran has its result withheldbestätigt 10.11.2026
• Changed the turn.start mod hook to also hear subagents’ and teammates’ turnsbestätigt 10.11.2026
• Added support for rewriting a workflow agent’s prompt and model in the agent.spawn mod hook, which could only refuse workflow agents beforebestätigt 10.11.2026
• Added the prompt.change mod hook, raised when the prompt box’s text changes without a typed editbestätigt 10.11.2026
• Added e.isReadOnly to a mod’s tool.check event, set when the tool call is read-onlybestätigt 10.11.2026
• only the permission dialog or a host’s callback can approve the ask, and it is refused with the mod’s reason where nobody can be askedbestätigt 10.11.2026
• to let admins opt SDK managedSettings (parent tier) into the policy mergebestätigt 10.11.2026
• use deniedMcpServers to keep it offbestätigt 10.11.2026
• a nested CLAUDE.md over the size limit now loads only its startbestätigt 10.11.2026
• from Anthropic’s first-party MCP servers at api.anthropic.com to reach Claude in full instead of being cut at 4,096 charactersbestätigt 10.11.2026
• to run its turn at xhigh effort on Opus 5.5, Sonnet 5.5, Haiku 5.5 and Fable 5.1bestätigt 10.11.2026
• waits for the real completion instead of fabricating resultsbestätigt 10.11.2026
• it does not change the model for server-side auto modebestätigt 10.11.2026
• now also available on Bedrock, Vertex, Foundry, Claude apps gateway and telemetry-disabled sessionsbestätigt 10.11.2026
• an Owner or Claude Tag admin can let Claude in Slack answer in channels shared with another organization; it’s off by defaultbestätigt 10.11.2026
• Deprecated fast mode for Opus 4.8 as of October 30, 2026; it will be removed on November 30, 2026.bestätigt 10.11.2026
• After removal, fast mode requests on Opus 4.8 return an error and do not fall back to standard Opus 4.8.bestätigt 10.11.2026
• settings fetched from Anthropic’s API now apply without an approval prompt, and a stored copy that doesn’t verify is refusedbestätigt 10.11.2026
• Changed Claude’s file tools to ask before editing the local copy of your organization’s server-managed settings, as they do for your settings filesbestätigt 10.11.2026
• Added CLAUDE_CODE_SUBAGENT_EFFORT to set the effort level subagents run at when nothing else names one, instead of inheriting the session’sbestätigt 10.11.2026
• which turns the ultrathink keyword offbestätigt 10.11.2026
• Changed ultrathink to run its turn at xhigh effort and show in a rainbow in the chat box, as in the terminalbestätigt 10.11.2026
• Changed subagents and teammates on a different model from the main conversation to use that model’s saved effort level or its default, instead of the session’s effort levelbestätigt 10.11.2026
• typed while Claude is working, it raises the rest of that turnbestätigt 10.11.2026
• Changed the ultrathink rainbow to appear only where the keyword can raise effort, and not when the word is quoted or part of a file pathbestätigt 10.11.2026
• a change made mid-task now says it may wait for your next message, and the note clears when it no longer appliesbestätigt 10.11.2026
• Added an effort level to the model picker when you create or edit a routine that runs in the cloudbestätigt 10.11.2026
• Changed model names and IDs to drop the [1m] suffix and its label on models where the suffix has no effect; the suffix is still acceptedbestätigt 10.11.2026
• Added the date a stored claude.ai sign-in ends to claude auth status (loginExpiresAt) and to claude -p JSON results (credential_expires_at)bestätigt 10.11.2026
• Added a one-line stderr warning to claude -p text output, starting 10 days before a stored claude.ai sign-in endsbestätigt 10.11.2026
• Changed subagents to stop instead of switching models whenbestätigt 10.11.2026
• Changed background subagents to move a shell command to the background after 3 minutes on all providers, so the prompt cache doesn’t expire while it runsbestätigt 10.11.2026
• Changed SendMessage so a message to a running background agent arrives earlier: its running command moves to the background, without being stoppedbestätigt 10.11.2026
• Changed agent tools: lists to take only the named MCP server’s toolsbestätigt 10.11.2026
• claude mcp serve never offers thembestätigt 10.11.2026
• row to /tasks that lists finished subagents, so you can reopen one and message it even after it leaves the list or you restartbestätigt 10.11.2026
• Hardened the Agent tool against indirect prompt injection via content a subagent readbestätigt 10.11.2026
• with the Bash sandbox on, skills in .claude/skills folders below the starting folder load only if you start Claude Code there or /cd therebestätigt 10.11.2026
• Changed compaction to send Claude the skill list again afterwards; /reload-skills now sends only new or changed skillsbestätigt 10.11.2026
• can register at most 1,024 hooks and 16,384 matcher characters per sessionbestätigt 10.11.2026
• a skill used in the main conversation whose hooks are kept out no longer gets its allowed-toolsbestätigt 10.11.2026
• refuse a ! shell command that contains raw control characters other than tab and newlinebestätigt 10.11.2026
• one whose file or folder name has hidden, direction-changing or blank-like charactersbestätigt 10.11.2026
• Fixed plan mode running a write (an MCP tool’s, a file’s, a shell command’s) that a PreToolUse hook or a mod’s tool.check hook allowedbestätigt 10.11.2026
• Fixed Claude carrying out an unapproved plan after a window reloadbestätigt 10.11.2026
• a step interrupted in plan mode now stops with a notice when the session comes back in another permission modebestätigt 10.11.2026
• Added a one-time note when plan mode won’t ask before most edits or commands because bypass permissions is availablebestätigt 10.11.2026
• Added CLAUDE_CODE_AUTO_MODE_MODEL to choose the model the client-side auto mode classifier usesbestätigt 10.11.2026
• Added classifier_verdict and permission_mode to OpenTelemetry tool_decision events, so auto mode classifier decisions can be told apart from permission rulesbestätigt 10.11.2026
• Added permissions.denyInsteadOfPrompting setting: a tool call that would wait on a permission prompt is denied and the turn continues, in terminal sessions, subagents and teammatesbestätigt 10.11.2026
• Added sandbox.allowNestedUserNamespaces so sandboxed commands on Linux can create user namespaces (Chromium, rootless podman); this lets them get around the sandbox’s file protectionsbestätigt 10.11.2026
• Changed auto mode in Cowork cloud sessions: a hook’s allow of an MCP tool call now counts as no answer, so the classifier or a permission prompt decidesbestätigt 10.11.2026
• Changed auto mode with telemetry or feature flags off, including behind LLM gateways: calls the server can’t check are asked about or denied, with no fallback to the built-in classifier; set CLAUDE_CODE_AUTO_MODE_SERVER=0 to keep using it (Bedrock, Vertex and Foundry are unchanged)bestätigt 10.11.2026
• Changed commands the settings exclude from the sandbox to ask before writing to its write-list paths while the sandbox is enabled but cannot run; the unsandboxed fallback is refused therebestätigt 10.11.2026
• Changed the dangerous rm and rmdir prompt in bypass permissions mode in the Desktop app and VS Code to deny the command at once with a rewrite hint, so unattended sessions keep goingbestätigt 10.11.2026
• Fixed project verify skills being rewritten on every session instead of only when a documented command changedbestätigt 10.11.2026
• Added CLAUDE_CODE_IDLE_COMPACT_MIN_TOKENS to change idle compaction’s 200K-token minimumbestätigt 10.11.2026
• a conversation too long for the new model is now summarized by the previous model firstbestätigt 10.11.2026
• Added the names of the tools called earlier to the compaction summary in the transcript view (Ctrl+O)bestätigt 10.11.2026
• like other Bash edits, it is no longer restored by /rewindbestätigt 10.11.2026
• Claude now sees the call and is told its outcome is unknownbestätigt 10.11.2026
• Added Claude Code cloud sessions and Remote Control permissions to Enterprise custom roles, so Owners can control access to each by rolebestätigt 10.11.2026
• Changed turning on Cloud sessions or Remote Control in admin settings to ask for confirmation and list the custom roles that have that permissionbestätigt 10.11.2026
• Changed deniedMcpServers URL entries to also block Anthropic-provided servers whose upstream URL matches, and connectors in cloud sessions and on self-hosted runners, where the session is given the connector’s own URL; otherwise deny a connector by namebestätigt 10.11.2026
• Added support for MCP tasks in the terminalbestätigt 10.11.2026
• Changed MCP tasks to work in the terminal without feature-flag fetching (Bedrock, Vertex, Foundry, LLM gateways, DISABLE_TELEMETRY); set CLAUDE_CODE_DISABLE_MCP_TASKS=1 to turn them offbestätigt 10.11.2026
• when an MCP server runs a tool call as a task, Claude Code moves it to the background right away and tells Claude the result when it finishesbestätigt 10.11.2026
• Added the CLAUDE_CODE_DISABLE_MCP_TASK_BACKGROUND environment variable to make Claude Code wait for an MCP task inside its tool call instead of moving it to the background right awaybestätigt 10.11.2026
• only plugins from an official Anthropic source or a source that managed settings vouch for keep that pre-approvalbestätigt 10.11.2026
• scripts pass --trust-workspace, which takes no value (put a PR number or base branch before it), or set CLAUDE_CODE_DISABLE_ULTRAREVIEW_TRUST_CHECK=1bestätigt 10.11.2026
• with CI set they now refuse an untrusted folder at oncebestätigt 10.11.2026
• Added CLAUDE_CODE_RETRY_WATCHDOG_MAX_USAGE_LIMIT_WAIT_MS so unattended retry mode (CLAUDE_CODE_RETRY_WATCHDOG) ends a request at a usage limit that resets too far away instead of waitingbestätigt 10.11.2026
• Added CLAUDE_CODE_RETRY_WATCHDOG_MAX_WAIT_MS to limit how long unattended retry mode (CLAUDE_CODE_RETRY_WATCHDOG) waits out 429 and 529 errorsbestätigt 10.11.2026
• Added CLAUDE_CODE_RETRY_WATCHDOG_USAGE_LIMIT_RECHECK_MS so unattended retry mode (CLAUDE_CODE_RETRY_WATCHDOG) notices a usage limit that was lifted before its resetbestätigt 10.11.2026
• Added a Bash editor with colors, line numbers and a Copy button to the setup script box in cloud environment settingsbestätigt 10.11.2026
• Added a note to /loop confirmations in cloud sessions that a timer stops with its container, with an offer of a loop that survivesbestätigt 10.11.2026
• Added a warning to /schedule when a new cloud scheduled task gets connectors you did not ask for, with how to remove thembestätigt 10.11.2026
• Changed self-paced /loop to end when a turn finishes without rescheduling, instead of scheduling a fallback wakeup; set CLAUDE_CODE_LOOP_KEEPALIVE=1 to keep the old behaviorbestätigt 10.11.2026
• Self-hosted runner: Changed checkout hooks to be stopped after 30 minutes by default, failing the session’s setup; if yours needs longer, raise --checkout-hook-timeout-min or set it to 0bestätigt 10.11.2026
• Fixed claude -p exiting with code 0 and no output when startup failed; it now prints the error and exits 1bestätigt 10.11.2026
• Fixed claude -p exiting 0 when a full disk or quota lost its transcript or ~/.claude.json writes; it now prints an error and, in text and json output, exits 1.bestätigt 10.11.2026
• Fixed claude -p exiting 0 in auto mode when its actions were refused because no safety check was available: text and JSON output now exit 1bestätigt 10.11.2026
• Fixed Claude deleting your most recently updated /team-onboarding guide when asked to delete one without naming whichbestätigt 10.11.2026
• Fixed Claude saying a /team-onboarding guide was deleted when the server refused the deletebestätigt 10.11.2026
• Fixed @-mentioned text files over 256KB being left out silently: Claude is now told the file’s size and to read it in portionsbestätigt 10.11.2026
• to run at most 24 at a time per session, sub-agents included, plus at most 12 that have waited longer than their event’s timeoutbestätigt 10.11.2026
• Improved artifact database writes: an update can now remove a single field instead of rewriting the whole documentbestätigt 10.11.2026
• Improved the error for an artifact database write refused at the database’s size limitbestätigt 10.11.2026
• Changed what Claude is told after safeguards stop a subagent: retry with a narrower prompt or tell you, instead of doing the work itselfbestätigt 10.11.2026
• Changed auto mode so that a project’s settings files or a plugin can no longer switch its server-side check back on once you have switched it off outside the projectbestätigt 10.11.2026
• Changed a session’s exact Bash or PowerShell ask rule to be refused, not asked about, when a broader deny rule covers the commandbestätigt 10.11.2026