claude-code v2.1.329
- Source
- github.com
- Author
- claude-code-publisher[bot]Top Viber
- Date

Context
The change in Claude Code 2.1.329 most likely to alter how an existing setup behaves concerns hooks in Code Mode. Claude Code is Anthropic's coding , which runs commands and edits files on a developer's machine. Hooks are checks that a user or administrator configures to run at set points, and a PreToolUse runs before each tool call and can block it. Code Mode, an experimental setting that is off by default, has Claude write JavaScript that makes several tool calls in one step instead of requesting them one at a time, as described in earlier coverage of its documentation. The Code Mode documentation says each call a script makes still goes through PreToolUse hooks under the tool's usual name.
Until this release, Code Mode was stricter than the rest of Claude Code when such a hook broke. The hooks reference says that on most events, when a hook fails or times out, Claude Code carries out the action anyway unless the hook sets `"onFailure": "block"`. The Code Mode documentation, as read on November 20, still describes the exception: a call from a script was refused when its PreToolUse hook could not start, was killed or returned invalid output, even with `onFailure` unset. According to the release notes, a PreToolUse hook that errors no longer refuses a script's tool call, matching the behavior outside Code Mode. A command, HTTP or tool hook that must refuse now needs `"onFailure": "block"`.
The effect falls on hooks used as guards. A hypothetical hook written to stop pushes to a main branch used to fail closed in Code Mode when it broke, and now lets the call through unless it carries that setting. The notes speak of a hook that errors; they do not say whether a hook that times out in Code Mode still refuses the call, which the documentation treats as a separate case.
Two other entries make hooks easier to see. The folder trust dialog now warns where a folder's hooks are declared, which matters because project hooks can sit in settings files such as `.claude/settings.json` that are committed to a repository. Claude Desktop and other non-interactive hosts now show a warning in the conversation when hooks in managed settings, the configuration an organization pushes to its developers, fail to load. Most of the release's many other entries fix bugs in Code Mode, MCP server connections, sign-in and the terminal interface.
- Claude Code v2.1.329 adds a trust-dialog warning that says where a folder's hooks are declared, and a conversation warning in Claude Desktop and other non-interactive hosts when hooks in managed settings fail to load.
- In Code Mode, a PreToolUse hook that errors no longer refuses a script's tool call, matching behavior outside Code Mode. A command, HTTP or MCP tool hook that must refuse now needs "onFailure": "block" set.
- The release removes the automatic retry added in 2.1.326 at the limit named in a max_ error. The notes say users behind a gateway whose model takes fewer output tokens should set CLAUDE_CODE_MAX_OUTPUT_TOKENS to that limit.
- Scheduled tasks saved by earlier versions mostly need scheduling again in new signed-in sessions, although --resume of their own session still runs them.
- /code-review given a pull request number now reads it with gh pr diff, or says when it can't, instead of sometimes silently reviewing the checked-out branch. With no target, it reviews the whole branch against the default branch on most models.
- AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
- agent hook — A script the harness runs automatically at a fixed point in an agent's loop — before a tool call, after an edit, on session start.
- MCP — The Model Context Protocol — an open standard that lets any AI assistant plug into any tool or data source without custom integration code.
- token — The chunk of text a model reads and writes in — roughly three-quarters of a word — and the unit AI usage is billed in.
Checking sign-in…
Loading comments…





