Claude Code 2.1.330: Enterprise Admin Console Sign-In, easel Chart Module, and Scheduled Tasks Extended to 30 Days

Anthropic's Claude Code 2.1.330 changelog lists Enterprise Admin Console sign-in, a built-in easel chart module, scheduled tasks of up to 30 days, and stricter approval for rm commands.

News date: About 5 min read

Contents
  1. 1. What changed in 2.1.330 compared with earlier versions?
  2. 2. Which plans and platforms can use the new features?
  3. 3. How does this affect team and self-hosted runner workflows?
  4. 4. FAQ

Anthropic published Claude Code 2.1.330 in its changelog, dated 2026-11-20. The release covers enterprise sign-in, a chart module, a longer limit for scheduled tasks and stricter safety approvals. It also changes the VSCode extension and self-hosted runners.

What changed in 2.1.330 compared with earlier versions?

Anthropic's changelog lists three additions. You can now sign in to organizations managed by the Enterprise Admin Console; after you run claude auth login, the organization's settings apply. There is a built-in easel chart module. And setting CLAUDE_CODE_FULL_SIGN_IN_LINK=1 prints the full sign-in link, for when the short link shows an error or a script needs to read values from it.

Among the behavior changes, the changelog says recurring scheduled tasks and /loop (a command that makes Claude repeat the same job) can now run up to 30 days, and they say when they end. Tasks created before this version still end after 7 days. Anthropic also restored the automatic retry when max_tokens (the limit on reply length) exceeds a model's or gateway's limit. A gateway is a service that routes requests to the model. Version 2.1.329 had removed this retry.

Selected behavior differences in Anthropic's changelog ("—" means the changelog does not state the earlier status)
ItemBefore2.1.330
Recurring scheduled tasks / /loopUp to 7 daysUp to 30 days while the session keeps running; older tasks remain at 7 days
rm commands (root, home, top-level folders, working folder)—More commands require approval; refused where nobody can answer
Self-hosted runner spawn-runner hooks—Can start while earlier hooks are still running, up to 4 concurrent per orchestrator by default, adjustable with --hook-concurrency
Bash file diffs (self-hosted runner)—Skipped when not needed, so transcripts may lack them; set bashEditDiffEnabled to true in managed settings to get every diff
deniedMcpServers name entries—Hide only the named claude.ai connector; to block all, deny its exact URL

Which plans and platforms can use the new features?

The changelog states a plan and platform scope only for easel: Pro, Max, and (where available) Team and Enterprise, on macOS and Linux. You can turn automatic charts off in /config. Other items, such as enterprise sign-in, list no plan or region details.

The changelog lists several additions to the VSCode extension: - /fork, which branches the conversation after the last reply. - /workflows, which opens an agent map showing running workflows. - Three settings: claudeCode.escapeStopsClaude, claudeCode.pinPrompt and claudeCode.contextUsageThreshold. - A prompt argument for the Open in Side Bar command. - Move up and Move down in the session group menu.

How does this affect team and self-hosted runner workflows?

A self-hosted runner is an environment where Claude Code jobs run on machines you manage. Hooks are commands that run automatically at specific moments. The changelog says self-hosted runners add --start-dir / SELF_HOSTED_RUNNER_START_DIR, which lets Claude start work in a folder of the repository. Settings and hooks still load as they would without it. Two new /metrics metrics measure seconds from a work order being issued to registration being accepted, and to the first session being handed over. If a checkout hook exits with status 108, the runner immediately stops starting that session and does not try other runners.

For plugins, the changelog says cloud and self-hosted runner sessions now show a notice when a plugin marketplace can't be fetched because its repository isn't one of the session's. Results from claude plugin install --json also gain commandPrinted and commandFailed fields. The changelog also lists model-related changes: - Unrecognized models whose names contain sonnet or haiku now use a shorter system prompt. Setting CLAUDE_CODE_SIMPLE_SYSTEM_PROMPT=0 gives all models the full version. - The model field of prompt hooks accepts aliases such as haiku and sonnet. - In terminal sessions signed in to a Claude apps gateway, subagent progress summaries are now written from short notes, using far fewer tokens.

FAQ

Does easel draw charts automatically? Can I turn it off?

Per the changelog, easel draws charts when asked and also appears below MCP metrics results. You can turn automatic charts off in /config.

Will existing 7-day scheduled tasks automatically become 30 days?

No. Per the changelog, tasks created before this version still end after 7 days. New tasks can run up to 30 days while the session keeps running.

What if the short sign-in link shows an error?

The changelog says you can set CLAUDE_CODE_FULL_SIGN_IN_LINK=1 to print the full sign-in link. This also helps when a script needs to read values from the link.

How does deniedMcpServers behave differently?

Per the changelog, name entries now hide only the named claude.ai connector, not other connectors at its URL. To block all of them, deny its exact URL.

Browse the latest news in this topic

Found this guide helpful? Share it with someone who needs it

Sources

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.

Lifestyle