claudekit / updates / claude-code-2-1-314
[ PATCH · ]

Claude Code 2.1.314

Adds `permissions.blockWritesOutsideWorkingDirectories`, which makes Write, Edit and NotebookEdit refuse paths outside the working directories in every permission mode, and idle compaction in VS Code and Claude Desktop's Code tab. Over 100 changes, including fixes for sessions failing turn after turn with "Invalid effort level" when an admin caps effort and `claude -p` exiting with code 0 and no output when startup failed.

Official announcement →

This article is a summary based on official documentation.

What changed

Claude Code 2.1.314 shipped on October 30, 2026. It adds a permission setting that blocks writes outside the working directories, IdP group routing for Claude apps gateway upstreams, and idle compaction in VS Code and Claude Desktop. Over 100 changes span effort caps, Bedrock and Foundry, TLS certificates, MCP sign-in, managed settings, sandbox and hooks, vim mode, the VS Code extension, cloud sessions and Claude Tag.

New features

  • permissions.blockWritesOutsideWorkingDirectories

    Depending on the permission mode, Claude could write to or edit files outside the working directories. With this setting, Write, Edit and NotebookEdit refuse paths outside the working directories in every permission mode.

  • groups for Claude apps gateway upstreams

    There was no way to limit an upstream to certain people. An optional groups list on an upstream now means only developers in the listed IdP groups are sent to that upstream, on failover too.

  • guardrail_trace for Bedrock upstreams

    When a request was blocked on a Claude apps gateway Bedrock upstream, it was hard to tell why. guardrail_trace logs which Amazon Bedrock guardrail filter fired on a request, without the matched text.

  • Idle compaction in VS Code and Claude Desktop’s Code tab

    Coming back to a long conversation after the prompt cache expired meant the whole conversation had to be processed again. Now a 200K+ token conversation left idle is summarized, and the summary is used on your next message if the prompt cache has expired.

  • [Cloud sessions] Include completed button

    When finished one-off tasks are the only ones hidden, the Scheduled tasks page shows an Include completed button, so you can show them without opening the Sort menu.

Key improvements

  • Lower memory use with many subagents

    In long headless and SDK turns that run many subagents, the parent session kept each subagent’s tool results until the turn ended. It no longer does.

  • Auto mode’s blocked calls and /permissions Recently blocked

    Messages for auto mode’s blocked calls now name the command, say Claude may retry and auto mode checks again, and point to /permissions to review. The /permissions review tab is now called Recently blocked, its footer says what Esc will do with marked rows, and it refuses approving a call you can’t read in full, including one the terminal is too short to show.

  • Shorter reasons in permission prompts

    A long reason could push the command being approved off the screen. A long reason now takes fewer rows.

  • Ctrl+R matches pasted text

    Prompt history search with Ctrl+R now matches text inside collapsed [Pasted text] pills.

  • Auto-titles for sessions started with a skill or command

    A session that starts with a skill or a prompt-based command such as /security-review now gets a title from it instead of staying untitled.

  • Faster --resume and /context

    --resume is faster and uses less memory on damaged transcripts, and /context is faster in sessions with ten thousand or more MCP tools.

  • /design-sync upload prompt

    The upload prompt now shows the folder, the project and the files it will upload or delete.

  • Bundled skill updates

    The /claude-api skill’s Datadog setup for API tracing now sends the dd-llmobs-enabled: true header, as the Console does, and its ant install instructions drop the outdated macOS xattr step. The bundled dataviz skill now answers donut and pie chart requests with ranked bars or a segmented bar that prints its values.

  • which in the Bash tool

    An inherited which alias or function could change what which did. The Bash tool’s shell now clears it and stops passing exported which/unsetenv functions to commands, so which is the system command or zsh builtin.

  • Hooks and the Monitor tool

    Hooks matching Bash didn’t run on shell commands started by the Monitor tool. They now do, and can block or ask, but not approve or rewrite those commands. Hook if conditions such as Monitor(tail *) now match the Monitor tool’s shell commands as Bash(tail *) matches Bash’s.

  • Retry on another model when only an older Opus is left

    When ANTHROPIC_DEFAULT_OPUS_MODEL, availableModels or deniedModels leaves only an older Opus model, or none, flagged requests now retry on another model instead of ending in a refusal.

  • Idle compaction in more setups

    Idle compaction now also runs for subscribers with telemetry turned off, and on Bedrock, Vertex, Foundry and LLM gateways once the one-hour prompt cache is on.

  • Asking before editing server-managed settings

    Claude’s file tools now ask before editing the local copy of your organization’s server-managed settings, as they do for your settings files.

  • [Cloud sessions] Starting despite a failing setup script

    Cloud sessions from the web, desktop and mobile apps failed to start when a setup script exited with an error. They now start anyway. Scheduled tasks that run in Anthropic’s cloud now load the plugins enabled on your claude.ai account.

Bug fixes

Effort and models

  • Sessions failing turn after turn with “Invalid effort level” when an organization’s admin caps a model’s effort — Claude Code now lowers the effort, retries, and says which level it uses.
  • Model IDs spelled with a dot behind an LLM gateway, such as claude-opus-4.8, being treated as an older model — which gave them the wrong context window, output limit, effort, thinking mode and cost.
  • Microsoft Foundry sessions failing when the resource has no deployment for the default Sonnet, Opus or Haiku model — Claude Code now falls back to an older model of that family.
  • The Bedrock and Vertex model upgrade offer coming back at every start when managed settings pin the model.
  • Model switching being refused on machines with managed plugin settings when one of your own --plugin-dir or --plugin-url plugins had a broken manifest or failed to download.

Cloud providers and networking

  • Bedrock sessions with many images failing every turn and /compact with “400 Input is too long.” — older images are now removed and the session compacts.
  • Requests failing when a gateway or cloud provider refuses one of Claude Code’s beta headers by name — the request is sent again without it.
  • Repeated refused requests behind a gateway that doesn’t accept tool search’s tool_reference blocks — the request is sent again once without tool search, and the session remembers.
  • TLS connections failing with UNABLE_TO_GET_ISSUER_CERT when the system certificate store holds a renewed root certificate that is cross-signed with its predecessor.
  • A Claude apps gateway Vertex upstream hanging for minutes when a Google token endpoint accepts the connection and never answers.

Managed settings and security

  • forceRemoteSettingsRefresh set only in remote managed settings no longer applying once the stored copy had expired — this let sessions start without policy when the server could not be reached.
  • An administrator’s TLS trust settings in managed settings not being applied in Claude Desktop sessions.
  • A settings file excluded with --setting-sources still supplying some settings, such as auto mode rules and saved permission consents.
  • Sandbox protections not applying to a worktree entered by claude -p --resume or --continue.
  • claude ultrareview and claude remote-control asking the workspace trust question in a CI job that has a terminal — with CI set they now refuse an untrusted folder at once. /ultrareview now also checks workspace trust before it runs git.
  • Auto mode reporting a call as blocked, and listing it in /permissions, when its safety check gave no verdict.

MCP and sign-in

  • MCP servers that also work without sign-in showing “needs authentication” after their stored sign-in stopped working — they are now used signed out.
  • claude mcp logout reporting success when the MCP sign-in was not removed, for example while the macOS keychain is locked over SSH.
  • Extra sign-in renewals when several sessions share an Anthropic profile login and the credentials lock is busy.
  • /upgrade opening the browser and starting a new login for plans billed through the App Store or Google Play — it now says where to change the plan.

Headless and automation

  • claude -p exiting with code 0 and no output when startup failed — it now prints the error and exits 1.
  • “Continue automatically at usage limit” continuing more than three times without a prompt from you when a task restarted itself between waits (a /goal check-in, a background or scheduled task).
  • Background and cloud sessions started with Ultracode turning it back on at their next start after it had been switched off.
  • A PreCompact hook with very large output making compaction fail or repeat — each hook’s output is now cut at 10,000 characters.
  • Background diff refreshes briefly taking the repository’s index lock, which could make a git command run at the same moment fail with “Unable to create ‘.git/index.lock’: File exists”.

Display and input

  • Markdown tables jumping between grid and stacked layouts on half-typed rows while streaming, and long tables falling back to raw pipes.
  • The /resume picker showing the text of CLAUDE.md or CLAUDE.local.md as a session’s title when the session only ran slash commands.
  • Pasting an image over SSH to a Linux machine doing nothing silently — the prompt now says to copy the file over and paste its path.
  • Claude being told only that a command was declined after you stopped it with Esc or Stop in a cloud session or the IDE — it now also hears that the command may have run.
  • Vim mode dw, dW, yw and yW carrying on past the end of the line — w and W now stop on empty lines. Also fixed: aw and iw taking the line break at the end of a line, and p over a character selection ignoring whether the yanked text was whole lines.
  • Pasted text and dragged-in files being lost in Claude’s question dialog unless the “Other” row had focus.

Skills and cleanup

  • ~/.claude/skills/.trash and ~/.claude/plugins/.trash growing without limit — each now holds about 1024 entries, whatever cleanupPeriodDays says. The skill file watcher also no longer checks every removed skill kept in .trash.
  • The Bash edit diff listing a whole file as changed when a CRLF file under git’s line-ending conversion was re-saved unchanged or had one line edited.

VS Code extension

  • “Clear authentication” for MCP servers reporting success when the saved sign-in could not be removed.
  • The voice dictation shortcut (Ctrl+D) doing nothing under non-Latin keyboard layouts such as Russian, Greek or Hebrew.
  • Missed one-time saved scheduled tasks being deleted without asking when the extension opened without a chat.
  • Session group changes that cannot be saved, such as a delete, failing silently — an error notification now appears.

Cloud sessions and Remote Control

  • claude remote-control sessions failing to start when the session service was slow for a moment — the worker’s registration is now retried with growing waits.
  • The permission mode picker staying on Auto after a running cloud or Remote Control session refused the switch — it now shows the mode in use and tells you.
  • WebFetch in Cowork cloud sessions silently dropping a PDF’s text past 100,000 characters — it now says how much was unread and where to read on.
  • Cloud sessions reporting uncommitted changes every turn after Claude moved a branch to a commit it had pushed from another worktree.

Claude Tag

  • @Claude !restart giving up after one failed workspace check in channels shared across Enterprise Grid workspaces — it now retries, then asks you to run it again.
  • Follow-up messages in Slack showing no sign of progress while Claude’s session waits for a self-hosted runner — Claude now posts a wait notice in the thread.
  • Claude sometimes not answering a tag in its own thread in Slack Connect channels where both organizations have installed Claude.

Notes

  • To keep Claude from editing files outside the working directories, turn on permissions.blockWritesOutsideWorkingDirectories. Write, Edit and NotebookEdit then refuse those paths in every permission mode.
  • If you set FORCE_HYPERLINK=false, the behavior changes: false (also no and off) used to turn hyperlinks on, and now turns them off.
  • If a hook allows MCP tool calls in Cowork cloud sessions, that allow now counts as no answer under auto mode, so the classifier or a permission prompt decides.
  • If you use CLAUDE_CODE_DISABLE_CFC_PROMPT in VS Code, it now also stops the extension adding the Claude in Chrome instructions to a session’s messages. The browser tools stay available.
  • On macOS, git grep in most commands that Claude runs now uses the C locale, to avoid a reported memory leak in macOS’s regex library on files that aren’t valid UTF-8.
  • This post covers the main items. See the GitHub release notes for the full list.