What changed
Claude Code 2.1.301 shipped on October 15, 2026. It adds features to unattended retry mode, OpenTelemetry, vim mode, the mod API and the VS Code extension, and improves switching to a model with a smaller context window. The fixes span worktree cleanup, model aliases and sign-in, the Grep, Glob and Read tools, skill arguments, compaction, terminal rendering and the self-hosted runner. The release is large, so the items below are a selection of the changes with the most user impact.
New features
-
Usage-limit wait cap for unattended retry mode
Unattended retry mode (
CLAUDE_CODE_RETRY_WATCHDOG) waited at a usage limit until it reset, however far away that was. WithCLAUDE_CODE_RETRY_WATCHDOG_MAX_USAGE_LIMIT_WAIT_MS, it ends a request at a usage limit that resets too far away instead of waiting. -
Classifier fields on OpenTelemetry
tool_decisioneventsclassifier_verdictandpermission_modeare added totool_decisionevents, so auto mode classifier decisions can be told apart from permission rules. WithOTEL_LOG_TOOL_DETAILS=1,classifier_categoryalso names the built-in rule a classifier block matched. -
OpenTelemetry tracing guidance in the
claude-apiskillThe
claude-apiskill now covers OpenTelemetry tracing, including setting up trace destinations with theantCLI. -
%motion in vim modeJump to the matching bracket, alone or with d/c/y.
-
rowson a mod’sInputA mod’s
Inputcan now be a text field of several lines in the terminal, where Enter submits and Shift+Enter or a backslash before Enter starts a new line. -
[VSCode] LaTeX math rendering
Claude’s responses now render math, including
$$and multi-line\[ … \]blocks and inline$…$and\(…\)math. -
[VSCode] “Get more usage” link
The usage-limit banner for Pro and Max plans in the VS Code and JetBrains panels now has a “Get more usage” link.
Key improvements
-
Switching to a model with a smaller context window
A conversation too long for the new model is now summarized by the previous model first. A related fix: switching could previously spend usage credits on that summary without asking first.
-
Rendering performance with long-running subagents
A subagent’s progress trail now updates incrementally instead of being reprocessed in full on every render.
-
Artifacts with sticky headers
Pages Claude writes now paint the band above a sticky header, so scrolled text no longer shows there in the mobile apps.
-
Trees above the question dialog in mods
In VS Code and the desktop and mobile apps, a tree drawn above the question dialog can now hold Markdown and up to 48 rows. The terminal keeps 12 plain rows.
Changed behavior
-
A mod hook’s
.catchhandlerIt now runs only when the hook throws or times out. Events raised beneath the hook’s own calls skip that hook without asking its
.catch. -
A mod’s
$.env.setIt resolves with
{}once the variable is set or unset, or with{ deny }saying why when the write is not made.
Bug fixes
Worktrees & git
- Quitting removing a worktree made by a
WorktreeCreatehook without asking when it may hold unpushed commits. /diffnot noticing a conflicted cherry-pick or revert in reftable repositories when one of its git checks timed out or failed.
Models & sign-in
- The
fable(andbest) model alias erroring on the Anthropic API when the newest Fable model is not available to your organization; Claude Code now retries once on the previous Fable model. - A misleading “Please run /login” error when a full disk keeps the login from being refreshed; the error now names the disk and says to delete files.
- Sign-in refresh timeouts and unexpected sign-outs after a short command such as
claude auth statusorclaude plugin listexited in the middle of a token refresh. - Switching to a model whose name contains
application-inference-profilerunning yourawsAuthRefreshcommand and delaying model-switch hooks when the provider is not Bedrock. - Conversations being compacted over and over, on a model you didn’t pick, behind an LLM gateway whose responses name a different model than the one requested.
Tools & skills
- Grep and Glob reporting no matches when the search program failed unexpectedly before returning any results; they now return an error saying the search failed.
- The Read tool reading binary files such as .xlsm, .parquet, .epub and .heic as text.
- Skills and commands with very many or very long argument names stalling for seconds when invoked; only the first 64 names, each up to 64 characters, are now filled in.
- Skills, commands and hooks using very large amounts of memory when their arguments fill many placeholders.
- Skills no longer being listed to the model after using
/rewind. - The /claude-api skill saying CMEK external-key endpoints aren’t available on Claude Platform on AWS.
Workflows, artifacts & permissions
- Workflow agents sometimes skipping their step, and the run still reporting success, after reading the harness note that their task text carries no user authority.
- Workflow scripts asking for permission or being refused when read, edited or re-run after a Bash
cdin the same session. - Artifact updates being refused where a PreToolUse hook covers Edit or Write; Claude now edits the file and then republishes instead of changing it in the publish call.
- A tool call cut off by a message sent mid-turn during its permission check being reported to the model as rejected by the user.
- The permission prompt for a plugin or skill hook’s
askdecision not showing which plugin or skill the hook comes from.
Vim mode & input
- Vim mode sometimes staying in INSERT after Escape on a slow or busy machine.
- Esc discarding a typed AskUserQuestion answer in vim mode: the free-text field now supports vim NORMAL mode, and Esc cancels only from NORMAL.
Compaction & sessions
- [Fullscreen]
/compactand auto-compact briefly freezing the UI while the visible transcript is rebuilt. - The compaction row in a resumed session offering ctrl+o for history that is not loaded; it now says where the earlier messages are saved, and also shows in fullscreen.
- The “new task? /clear to save tokens” hint coming back right after you interrupt a response.
Terminal display
- Stray characters such as
>0qcovering the folder trust prompt, and>4mbefore the shell prompt at exit, in terminals built on VTE older than 0.53 such as MATE Terminal on RHEL 8. - Stray
GOKtext appearing in the prompt at startup in some versions of iTerm2. - A Markdown table with only a header row drawing nothing when too wide for the terminal, including in plans on the plan approval card.
- Dim text on the dark ANSI theme and 16-color terminals: now faint inside filled boxes, not faint under selection and search highlights, and list descriptions dimmed like other secondary text.
Mods
- A mod’s
$.model.forkreturningnothing-to-forkwhen called before the first response of a new Claude Code process had finished, as inclaude -pruns and cloud sessions.
Self-hosted runner
- Sessions started from Slack being unable to find or attach a repository when managed settings restrict MCP servers; use
deniedMcpServersto keep those tools off. - Sessions being reported as having a deleted source branch, and not retried, when an unrelated setup or session failure quoted git’s missing-ref wording.
- The log promising a retry of a failed or declined session release for a session that is ending; repeated failures now show a count.
Windows
- Stopped background PowerShell commands missing the closing
[killed]line in their output file.
VSCode
- A conversation that crashes or freezes the panel being reopened on every reload.
- The chat panel slowing down during long thinking: it no longer re-parses the whole thinking text on every update.
- Diagnostics showing “[object Object]” instead of the rule code (e.g. MD013) for some language servers.
- The native dictation audio module being loaded to check microphone permission where dictation is not offered (non-claude.ai sign-ins, remote windows).
Notes
- If you run unattended jobs with
CLAUDE_CODE_RETRY_WATCHDOG, setCLAUDE_CODE_RETRY_WATCHDOG_MAX_USAGE_LIMIT_WAIT_MSto end requests at a usage limit that resets too far away. - If you analyze permission decisions with OpenTelemetry,
classifier_verdictandpermission_modeontool_decisionevents separate auto mode classifier decisions from permission rules.classifier_categoryrequiresOTEL_LOG_TOOL_DETAILS=1. - If you write mods, when a hook’s
.catchruns and what$.env.setresolves with have changed. Check any code that relies on the old behavior. - If you create worktrees with a
WorktreeCreatehook, quitting no longer removes one that may hold unpushed commits without asking. - If the
fableorbestalias errored for you, Claude Code now retries once on the previous Fable model when the newest one isn’t available to your organization.