Claude Code 2.1.329: Code Mode Works With Telemetry, /tasks Lists Scheduled Tasks, macOS Computer Use Gets Stricter Safety Rules
Anthropic's changelog lists Claude Code 2.1.329, covering Code Mode behavior changes, new diagnostics in /tasks and /mcp, /code-review changes, and new safety rules for macOS computer use.
News date: About 7 min read
Contents
On 2026-11-19, Anthropic listed version 2.1.329 in the Claude Code changelog. The update centers on three areas: behavior changes to Code Mode (an execution mode in which Claude writes a script and the script calls tools), new diagnostics in commands such as /tasks and /mcp, and new safety rules for computer use on macOS (the feature that lets Claude operate apps on your computer). The changelog lists these as ordinary version changes; it does not label them preview, beta or waitlist, and it mentions no plan or region restrictions. If you rely on hooks (custom commands that run automatically at specific moments), scheduled tasks or managed settings (settings pushed centrally by an administrator), behavior after upgrading may differ from before.
What changed in 2.1.329?
According to Anthropic's changelog, /tasks now lists this session's scheduled tasks, /loop wakeups, the goal and this conversation's forks, and pressing x stops one after asking. The /tasks header and the footer's background count now show one number covering shells, monitors, MCP calls, agents, workflows and scheduled tasks.
For diagnostics, the /mcp server page gains a "Server message" row. MCP (Model Context Protocol) is the protocol that connects Claude Code to external tools and services; when a remote HTTP MCP server rejects a connection, this row shows the server's JSON-RPC error message verbatim. /status gains a "Marketplace auto-update" row indicating which managed marketplaces will not auto-update and why. The trust dialog also adds a warning explaining where a folder's hooks are declared.
For code review, a /code-review with no target now reviews the whole branch against the default branch on most models, and a bug where multi-commit branches lost diffs is fixed. /code-review with a PR number and --fix, and /code-review given a PR URL, now stop without reviewing if gh pr diff is denied or fails.
How do Code Mode and hooks behave differently?
OpenTelemetry is an open standard for collecting monitoring data such as events and metrics. Anthropic's changelog says Code Mode can now be turned on when OpenTelemetry events and metrics are enabled, with an event emitted for each tool call a script makes, but Code Mode stays off when tracing is enabled. Code Mode also starts when telemetry, nonessential traffic or feature flags are turned off, and admins can keep it off in managed settings.
Another notable adjustment concerns the PreToolUse hook (a hook that runs before a tool is called): a failing PreToolUse hook in Code Mode no longer denies the script's tool call, matching behavior outside Code Mode. Command, HTTP or MCP tool hooks that must deny can set "onFailure": "block". In addition, Code Mode now rejects reading line ranges over 256 MiB (which previously could consume several GB of memory), fixes timeouts when reading very long lines, and runs scripts that declare their own log, preview or sleep.
| Item | Change described in the changelog |
|---|---|
| Code Mode and OpenTelemetry | Can be turned on when events and metrics are enabled; tracing still keeps it off |
| Code Mode with telemetry/feature flags off | Also starts; admins can keep it off in managed settings |
| Failing PreToolUse hook | No longer denies the script's tool call; set "onFailure": "block" to deny |
| Reading line ranges | Ranges over 256 MiB are rejected |
| /code-review (no target) | On most models, reviews the whole branch against the default branch |
| auto mode | Runs git status less often; git commands that discard changes may be blocked until the working tree is checked |
| macOS computer use | Applies newer safety rules; app launches give up after 30 seconds |
Which changes may affect existing scheduled tasks and environments?
Anthropic's changelog says that in a new signed-in session, most scheduled tasks saved by older versions need to be rescheduled, although resuming their original session with --resume still runs them. "Exit and stop tasks" ends this session's scheduled tasks so they no longer run on resume, unless you exit while Claude is responding.
Computer use on macOS is updated to newer safety rules, so some actions are denied, some apps have less access, and app launches give up if not finished within 30 seconds. auto mode runs git status less often, giving most requests a shorter wait in large repositories.
For teams using a gateway, the changelog explains that 2.1.326's automatic retry at the limit named in a max_tokens error has been removed; if the model behind your gateway can output fewer tokens, set CLAUDE_CODE_MAX_OUTPUT_TOKENS to that limit. For self-hosted runners, a second-hostname copy of the same owner/name goes into a folder named after the hostname under _repos/clones/ in the base folder; repositories whose address path contains special characters are cloned again into a new folder, and rules and volumes that name the old folder no longer apply.
What is different for VSCode users?
Anthropic's changelog lists several VSCode changes: after you decline the trust prompt, a Review trust line appears above the message box so you can reopen the prompt without reloading the window; in permission prompts, pressing Down stops at the last item instead of wrapping back to Yes; and command output shows hidden characters and control sequences other than plain colors as symbols. Also, some teammate panes now launch through /bin/sh, so they may not read shell startup files.
Frequently asked questions
Is 2.1.329 a preview or limited to certain plans?
Anthropic's changelog only lists the changes in 2.1.329. It mentions no preview, beta or waitlist and states no plan or region restrictions. For actual availability, refer to Anthropic's official changelog.
Will a tool call still be blocked when a hook errors in Code Mode?
According to Anthropic's changelog, a failing PreToolUse hook in Code Mode no longer denies the script's tool call. If a command, HTTP or MCP tool hook must deny, set "onFailure": "block" on that hook.
Will previously saved scheduled tasks still run?
Anthropic's changelog says that in a new signed-in session, most scheduled tasks saved by older versions need to be rescheduled; however, resuming their original session with --resume still runs them.
What should I do if the model behind my gateway outputs fewer tokens?
Anthropic's changelog explains that 2.1.326's automatic retry at the limit named in a max_tokens error has been removed, and advises setting CLAUDE_CODE_MAX_OUTPUT_TOKENS to that model's limit.
Why might auto mode block a git command?
Per Anthropic's changelog, auto mode now runs git status less often, so most requests wait less in large repositories. In return, git commands that discard changes may be blocked until the working tree has been checked.
Browse the latest news in this topic
Found this guide helpful? Share it with someone who needs it
Kimi Code CLI v2.7.0 Adds a Native Amazon Bedrock Provider: Use Your AWS Credentials Instead of a Separate API Key
AWS says Moonshot AI added a native bedrock provider in Kimi Code CLI v2.7.0, letting you call Amazon Bedrock models with the AWS credential chain (including named profiles and SSO) without managing separate Bedrock API keys.
AWS Rounds Up 13 SageMaker Inference Features: Managed NVIDIA Dynamo on HyperPod, MIG GPU Partitioning, Cross-Region Routing
BMW Group Shares How It Governs AI Agents: 12,500+ Developers, Three Autonomy Levels and a Four-Stage CI/CD Gate
Amazon SageMaker AI Now Monitors vLLM GPU Inference Endpoints by Default: Latency Drift, Idle GPUs and Missed Optimizations
Sources
- 2.1.329Checked:
Compiled from the references and verification dates listed above; fares, schedules, product features, and terms can change, so please confirm the latest details on the relevant official website before travelling, booking, or purchasing.