claudekit / updates / claude-code-2-1-321
[ PATCH · ]

Claude Code 2.1.321

Adds Claude Fable 5.2 (`claude-fable-5-2`), now the default Fable model, and a `--settings` option for `claude remote-control`. Over 110 changes, including fixes for MCP OAuth requests sending secrets such as refresh tokens to another host on a redirect, and a read-only Bash command running outside the sandbox without a permission prompt.

Official announcement →

This article is a summary based on official documentation.

What changed

Claude Code 2.1.321 shipped on November 9, 2026. The headline change is Claude Fable 5.2, now the default Fable model, and saved Fable 5.1 picks move to fable. The release also changes how mods are permitted, how managed extraKnownMarketplaces work and how Claude apps gateway policies add up. Fixes make up more than half of the release, including MCP OAuth secrets sent to another host and Bash sandbox and permission-check gaps.

New features

  • Claude Fable 5.2 (claude-fable-5-2)

    Claude Fable 5.2 is now the default Fable model, with 1M context at $8/$40 per Mtok and $0.20/Mtok cache reads.

  • --settings <file> for claude remote-control

    Applies a settings file to every session claude remote-control starts. If the file changes, no new session starts until you restart claude remote-control.

  • [VSCode] Auth token and API key warning

    A warning now appears above the prompt box when both an auth token and an API key are set, matching the terminal’s startup warning.

Key improvements

  • Saved Fable 5.1 picks move to fable

    On Pro and Max plans this happens at once; on Team and Enterprise plans, once your organization allows it. Use /model claude-fable-5-1 to switch back.

  • /model picker and fast mode labels

    “More models” is now an unnumbered row below every other, and older models open under it without renumbering the rest. The fast mode labels in /fast, /config and /model name your current model when it supports fast mode, and “(Opus only)” was removed from the VS Code row.

  • Mods

    • Mods Claude writes now turn on without asking in auto and bypass modes, even with managed allowUnsandboxedCommands: false or allowManagedDomainsOnly, and don’t load in dontAsk mode.
    • Mods’ own tool calls ($.tool.call, $.mcp.call) raise no permission prompt, even under your own ask rules or in dontAsk and -p runs. Deny rules and organization policy still apply.
    • Headless sessions reload mods in .claude/skills when their files change. Set CLAUDE_CODE_PLUGIN_DIR_WATCH=0 to turn this off.
  • Permission and security changes

    • cp and mv spelled by a path, such as /bin/cp, get the same path checks as cp and mv.
    • Nested skill loading: .claude/skills folders inside ~/.claude/plugins no longer load when Claude works on files there.
    • Remote Control sessions keep the folder they were started in. A connected client can no longer move a session to another folder.
    • In Cowork cloud sessions, a PostToolUse hook can no longer replace the answer to AskUserQuestion or the plan in ExitPlanMode’s result. Claude reads the originals.
    • AskUserQuestion in claude -p and SDK sessions denies a host’s permission answer that sets the undocumented afkTimeoutMs input field.
    • Claude in Chrome tools that an app running Claude Code provides, such as the Claude desktop app’s Code tab, follow your organization’s Claude in Chrome policy.
    • The account email check refuses, without asking, only a shell command that holds your email and names a User-Agent, so other commands and scheduled runs no longer stop at its prompt.
  • Managed extraKnownMarketplaces

    A declared name can’t be added from another source, an entry from elsewhere is hidden, and plugins the declared marketplace lists are reinstalled from it. On a take-over from an undeclared source, saved data is set aside for cleanupPeriodDays, and options and secrets until claude plugin restore-set-aside.

  • Faster install and startup

    claude install and the install scripts finish 2 to 4 seconds sooner. Startup is faster thanks to loading trusted TLS certificates faster. On Windows, Claude Code finds programs on PATH without starting a helper process, and with a standard Git for Windows install it no longer checks which git.exe to run on every launch.

  • How Claude Code finds programs on Windows

    For programs it starts itself, such as git, Claude Code skips relative and UNC PATH entries and links onto a network share, and picks up PATH changes from settings.

  • Context savings

    Background command notifications and TaskStop results show Claude only the start and end of a command over 1,000 characters. In sessions with more images than fit in one request, Claude is told whether it has already seen a dropped image, and repeat reads of an unchanged image are limited.

  • Background subagent reports

    In auto mode in headless sessions, the desktop app and VS Code, reports from background subagents that wait together are answered in one model call, as in the terminal.

  • Remote Control

    A Remote Control session whose process crashed picks up a new message sooner. claude remote-control’s messages for an unanswered first-run question each say what to do next. A Claude Desktop Code tab session can now send 20 messages to your other sessions, up from 10, before it waits for you to type again.

  • Claude apps gateway

    • Spend limits on a busy replica skip fewer checks and record spend sooner.
    • Sign-in limit messages in Claude Code and on the /device page no longer name gateway settings; the gateway’s own log still names them.
    • Error messages show the gateway’s address without any user name or password.
    • The spend limit error shows the reset in your local time, such as “resets 4pm (America/Los_Angeles)”.
    • With policy_mode: layered, managed MCP servers add up over every matching policy, match: {} included, so a gateway already on it serves more after the upgrade.
    • Bedrock, Vertex and Foundry upstreams ignore the ANTHROPIC_*_BASE_URL variables in the gateway’s environment. Set base_url on the upstream to use another endpoint.
  • Artifacts

    When an artifact is in another organization, /login is offered only if it’s one of yours, and Claude says where a copy may be after an account move. Artifact design guidance treats Inter and Space Grotesk as good default typefaces. The bundled dataviz skill suggests a list of meters, one row per item, for several ratios that each have their own limit.

  • Ultra in plan wording

    The session-feedback message and the allowedProviders setting description name Ultra alongside Pro and Max, and the forceLoginMethod description covers every paid Claude plan. The “Upgrade to Max” options now say “Upgrade to Max or Ultra”.

  • Other changes

    • The agent view’s row for a claude --bg --worktree session that couldn’t set up its worktree shows the error from its beginning, not an internal tag or its last 200 characters.
    • Self-hosted runner: an on-demand runner that the server refuses because it is not needed (session ended or paused, work order superseded) exits 0 instead of 1.
    • OpenTelemetry hook events in non-interactive sessions leave two built-in hooks out of num_hooks and num_success after Bash and PowerShell calls.
    • The Cyber Verification Program sentence on a cyber safeguard block uses the same words as claude.ai.

Bug fixes

Permissions and security

  • MCP OAuth requests sending secrets such as refresh tokens to another host when a server redirected them.
  • A read-only Bash command running outside the sandbox without a permission prompt.
  • Bash permission checks judging a file write from the wrong folder and not prompting.
  • Deny rules and protected-file prompts not applying to the file that a cp or mv into a directory writes.
  • Bash reads such as grep -r foo "$PWD/src" asking for permission when the same command with the path typed out runs without asking.
  • A sent file showing under a different name in the Claude apps when its name hid text-direction or control characters — Claude now refuses to send such a file.
  • Skill and command names holding a Bamum letter that some fonts draw as blank — they are now refused.
  • PreToolUse mcp_tool hooks on stdio and HTTP MCP servers being skipped when their answer was over 16 MiB — the tool call is now refused.
  • [VSCode] A plugin’s permission request showing a name with hidden characters — it now reads “a plugin”.

File edits and notebooks

  • Edit changing the first of several matches, replacing a whole file on an empty old_string, or editing a notebook as text, on network paths (UNC shares, WSL folders, /net mounts).
  • NotebookEdit deleting or overwriting a notebook’s first cell when the call named no cell, on network paths.
  • NotebookEdit reporting success to hooks and telemetry when the edit failed, for example when the notebook changed while the permission prompt was open.
  • High memory use in long sessions that repeatedly read single lines from large files.

Bash, tools and session context

  • Bash commands sometimes running in the wrong folder when Claude sent several at once and an earlier one changed folders.
  • Bash commands coming back empty for zsh users whose .zshrc sets KSH_OPTION_PRINT.
  • claude -p sometimes telling Claude that background commands get 30 minutes when they are stopped after 10.
  • pkill, pgrep and pidwait in cloud sessions matching the session’s own parent processes.
  • WebFetch passing “No response from model” to Claude as page content when the model reading the page returned no text — the call now fails with a clear error.
  • Context usage estimates overcounting tool definitions when exact token counts are unavailable.
  • The Artifact tool telling Claude to add filters when an artifact’s database collection is too large to sort — it now says to read the collection page by page and sort the results.
  • The git status given to Claude at session start reading as clean when git status failed (it now reads as unavailable), and showing the first listed file as staged when its changes were not staged.

Settings, models and display

  • /config showing Manual as the default permission mode when new sessions start in auto mode — pressing Space on “Auto (default)” now saves Don’t Ask, not Plan.
  • The “Draws from usage credits” note showing after a switch to Fable on Enterprise seats whose plan includes Fable.
  • The fallback model sometimes declining a request or mentioning a stopped response when the first model was blocked before replying.
  • Fast mode losing replies, switching models or retrying for minutes when fast serving is rate-limited, overloaded or failing — it now tries standard speed on the same model first.
  • Prompt suggestions making extra API requests when the model replied with a tool call.
  • A rare crash on permission prompts and background tasks.
  • A command’s panel sometimes not coming back after a dialog closed, leaving no panel or prompt on screen until the next redraw.
  • Fullscreen sessions with some plugins ending with an error when a tool result that had got shorter scrolled off the top of the screen.
  • Screen reader mode delaying the prompt at startup when the system clock moved back, turning off when you return to a conversation from agent view, and taking as yes a stray “y” at the end of text typed as a y/n question opened.
  • /desktop on macOS saying Claude Desktop is not installed when it is in ~/Applications.
  • macOS: “Unable to read managed policy settings” at startup when a valid MDM profile was slow to read — the read is retried once, and a timeout is now reported as a timeout.

Hooks

  • Windows: FileChanged hooks stopping for paths set through watchPaths after the first shell command, when the session’s working directory had a lowercase drive letter.
  • A large hook output pointing Claude to a file that held another hook’s output from the same tool call.
  • Apps that embed Claude Code getting their own PreToolUse hook’s ask reason instead of another hook’s when both asked.
  • A PreToolUse hook’s ask being lost when another hook’s defer finished first and could not be honored.

MCP, plugins and skills

  • claude mcp login printing a SEP-2352 issuer warning each time you sign in again to a server that Claude Code registered with.
  • An MCP server sometimes showing as connected with no tools after it was reconnected or the session’s directory changed while its tools were still loading.
  • A blank screen or no answer at startup, and empty claude mcp list output, when a stored MCP sign-in record was null.
  • A hang or out-of-memory crash when a connector in a cloud session lists resources with very long names.
  • A project’s MCP server being disconnected after the first message in a Claude Desktop Code tab session when the desktop app has an MCP server or extension of the same name.
  • Plugin auto-updates and /plugin marketplace updates downloading a plugin again for each worktree or folder of a repository.
  • --plugin-dir . and --plugin-dir ./builtin writing generated type files for Claude Code’s built-in plugins into a builtin folder.
  • Project skills, commands and agents staying missing after their folder was removed and came back mid-session, such as with git stash, including after entering a worktree.
  • Project skills, commands and agents disappearing from the menu in SDK and desktop sessions after Claude entered a worktree.

Subagents, background agents and workflows

  • Auto mode losing a background subagent’s report that had started a turn when the conversation was compacted before Claude answered it.
  • One request carrying far more background agent reports than intended after Claude was stopped while several were waiting.
  • Sub-agent progress summaries running a request to its end after the model replied with a tool call.
  • Background agents being reported as stopped by the user, and refused a resume, when a service interrupted a cloud session and nobody pressed Stop.
  • Claude not being told when you stop or pause a running workflow, so it no longer says the workflow is still running.
  • Agent view erasing another agent view’s saved input text when it restarted for an update with nothing typed.
  • claude agents ignoring --permission-mode, --dangerously-skip-permissions, --model, --effort or --agent typed before agents when a flag such as --verbose was also on the line.

Network, sign-in and saving

  • A turn ending with “API Error: EPROTO: The TLS handshake failed” when the connection broke partway through a reply — the partial reply is kept or the request is retried, as before 2.1.302.
  • Sign-in, token refresh and logout waiting indefinitely when something on the network accepts the connection but never answers the TLS handshake.
  • New messages not being saved, and claude -p, SDK and VS Code sessions hanging, when memory ran out while the transcript was being read.
  • claude -p on Linux exiting 0 in silence when a full NFS export or a used-up NFS quota lost transcript lines — it now prints an error and, in text and json output, exits 1.
  • Files Claude sends from cloud and Remote Control sessions failing at once when an upload gets an HTTP 507 from a proxy — the upload now gets a second try.

Remote Control, cloud sessions and Claude Desktop

  • /compact showing twice in Remote Control on web and mobile.
  • Remote Control showing as off with “login expired” in the desktop app while its connection was still up, because the app was briefly unable to hand over a sign-in token.
  • A Remote Control message sent with “Queue for later” reaching Claude in the middle of the running turn.
  • Remote Control showing as connected after its connection closed for good while the session’s history was still uploading.
  • claude remote-control accepting a y typed without Enter at its first-run question when the terminal’s input ended.
  • Remote sessions giving no reply after a restart when a UserPromptSubmit hook blocked a replayed permission answer — these hooks no longer run on that replay.
  • A restarted cloud session dropping a message with a large image when the restart came before Claude’s first reply to it, and leaving a subagent’s message to the main conversation unanswered.
  • Cloud sessions telling Claude that a host is probably not allowed when it is allowed and the network proxy could not connect to it.
  • Messages from a Claude Desktop Code tab session to your other sessions staying paused after you typed while Claude was still working.
  • OpenTelemetry repository attributes (vcs.*) missing from Claude Desktop sessions that start on a pre-started process.
  • Claude in Chrome and computer use server processes staying alive after their parent exits behind a proxy that stalls connections.
  • Claude Desktop not showing spend limit warnings behind a Claude apps gateway whose policy has no code: rules.
  • Claude apps gateway refusing to start when a desktop setting holds a field newer than the gateway, such as a managed MCP server’s oauth.omitResource — it now warns and serves it.

Notes

  • If you saved a Fable 5.1 pick, it now moves to fable (Fable 5.2 by default). On Team and Enterprise plans this waits for your organization to allow it. Use /model claude-fable-5-1 to switch back.
  • Mods Claude writes now turn on without asking in auto and bypass modes, and mods’ own tool calls raise no permission prompt, even under your own ask rules. Use deny rules or organization policy for anything they must not do. Mods Claude writes don’t load in dontAsk mode.
  • Remote Control sessions now stay in the folder they were started in. A connected client can’t move them.
  • .claude/skills folders inside ~/.claude/plugins no longer load when Claude works on files there.
  • If you operate a Claude apps gateway, policy_mode: layered now adds up managed MCP servers over every matching policy, so it serves more after the upgrade. Bedrock, Vertex and Foundry upstreams ignore ANTHROPIC_*_BASE_URL; set base_url on the upstream instead.
  • If you use managed extraKnownMarketplaces, a declared name can’t be taken over from another source, and plugins the declared marketplace lists are reinstalled from it.
  • If a script checks exit codes, claude -p on Linux that lost transcript lines to a full NFS export or quota now exits 1, and an on-demand self-hosted runner refused as not needed now exits 0.
  • This post covers the main items. See the GitHub release notes for the full list.