Claude Code 2.1.330: Built-in Charts, Longer Scheduled Tasks, and Safer File Deletion
Claude Code 2.1.330 ships easel, a built-in chart-drawing tool for Pro, Max, and eligible Team and Enterprise users on macOS and Linux, alongside 30-day scheduled task windows, safer rm command handling, and fixes for freezing, data loss, and stats timezone bugs. The update also adds enterprise Admin Console sign-in and several VSCode quality-of-life settings.
Claude Code Gets Built-in Charts, Safer Commands, and Longer Automation Windows
Anthropichas shipped version 2.1.330 of Claude Code, one of its most feature-dense releases in recent memory. According to the Claude Code changelog, this release touches everything from a brand-new built-in chart-drawing tool to safer handling of destructive terminal commands, longer scheduling windows for automated tasks, and a wave of stability fixes that address freezing, session corruption, and data loss bugs. If you use Claude Code for professional work — whether on a Pro plan, a corporate Team plan, or an Enterprise deployment — this update changes several things you will notice day to day.
The Headline Addition: Easel, a Built-in Chart Tool
The single most visible new feature in 2.1.330 is easel, a built-in mod that draws charts when you ask Claude a question that calls for one — or automatically when Claude Code surfaces metrics from an MCP (Model Context Protocol) server. The Claude Code changelog confirms that easel is available on Pro, Max, and, where available, Team and Enterprise plans, on macOS and Linux. Windows users do not get it in this release.
In plain language: you no longer need to install a separate charting plugin or export data to Excel to get a visual. Ask Claude Code to summarise your query results, and if the output lends itself to a chart, easel draws one inline. If you find the automatic charts distracting, the changelog notes you can turn them off in /config.
A Scenario: A Finance Team in Pune
Picture a small finance team at a mid-sized manufacturing company in Pune. They use Claude Code on a Team plan to query their ERP’s database, summarise monthly cost variances, and draft reports for the CFO. Before 2.1.330, the analyst running Claude Code would get a table of numbers in the terminal and would then paste those into a spreadsheet to build a bar chart manually — a step that could take fifteen to twenty minutes per report cycle.
With easel active, when Claude Code returns the MCP metrics results from that database query, it now draws a chart automatically. The analyst sees the cost-variance trend visualised in the same session, without switching tools. If the chart adds noise rather than clarity for a particular query, she can toggle automatic charts off in /config and ask for them on demand instead. That is a small but real time saving compounded over dozens of reports per month.
Scheduled Tasks Now Run for 30 Days
Another meaningful change in 2.1.330 is the extension of recurring scheduled tasks and /loop sessions. Previously, the Claude Code changelog states, these ran for up to seven days while the session stayed active. The new limit is 30 days. The changelog also adds that Claude will now tell you when a scheduled task is about to end, rather than stopping silently.
This matters for anyone who uses Claude Code to run recurring automations — nightly data pulls, weekly report generation, periodic code review sweeps — where a seven-day ceiling forced manual restarts. Note that tasks created before this release still terminate at seven days; the 30-day window applies only to tasks started after updating to 2.1.330.
Safer Handling of Destructive Commands
One of the more quietly important changes in this release is a tightening of how Claude Code handles rm commands aimed at sensitive locations. The changelog states that more rm commands directed at the root directory, the home directory, a top-level folder, or the working folder will now ask for your approval before running. In situations where no one can answer — such as a fully automated pipeline — the command is refused outright.
This is a guardrail against a class of accidents where an AI agent, following a chain of instructions, could wipe out directories that are difficult or impossible to restore. For non-technical professionals who run Claude Code in automated or semi-automated modes, this change reduces the risk of a runaway script causing serious damage to a project folder.
Enterprise Sign-in Now Works with the Admin Console
For organisations using Anthropic’s Enterprise Admin Console, 2.1.330 adds direct sign-in support. According to the Claude Code changelog, after running claude auth login, Claude Code will apply the organisation’s settings and use its configured cloud provider or gateway automatically. This simplifies onboarding for enterprise teams where IT has already set up centralised policy and cloud routing — employees sign in once and the right configuration follows them.
VSCode Gets Several Quality-of-Life Additions
The VSCode extension picks up a focused set of improvements in this release:
/forklets you branch a conversation after its last reply, so you can explore an alternative direction without losing the original thread./workflowsopens the agent map showing running workflows, giving you a visual overview of what Claude Code is doing in parallel.claudeCode.escapeStops Claudeis a new setting that lets you decide whether pressing Escape stops Claude’s current action — useful if you have a habit of pressing Escape to dismiss UI elements and don’t want that to interrupt a long-running task.claudeCode.pinPromptcontrols whether your message input stays pinned to the bottom of the panel or scrolls away with the conversation.claudeCode.contextUsageThresholdlets you set when the context usage indicator appears — handy for teams that want an early warning before hitting context limits.
These are configurability improvements rather than fundamentally new capabilities, but they address friction points that frequent users of the VSCode integration will recognise immediately.
A Long List of Bug Fixes
The bulk of the 2.1.330 changelog is a substantial list of fixes. Some that are likely to affect professional users directly:
- Claude Code freezing after compacting a very long conversation — this fix alone will be a relief to anyone who works in extended sessions that require compaction to stay within context limits.
- WebFetch freezing or returning empty results — the tool that fetches web pages for Claude to read was hanging on certain pages; this is now resolved.
- User settings being lost when multiple Claude Code processes saved simultaneously — a data-loss bug relevant to anyone who runs Claude Code in several terminals or automated pipelines at once.
/rewindrestoring an older file version incorrectly — this affected users who queued prompts or ran stacked skills in quick succession./statscounting activity in UTC instead of local time — an issue that made usage statistics confusing for users outside UTC, including everyone in India operating at IST (UTC+5:30).- Jupyter notebooks with a byte order mark being reported as invalid JSON — directly relevant to data analysts who work with notebooks.
- The built-in chart tool (easel) failing in Code Mode — it was broken in the same release that introduced it in certain modes; this is now fixed.
Limitations and Honest Tradeoffs
Several things in this release are worth noting as incomplete or restricted:
What to Watch For Next
The cadence of Claude Code releases has been rapid, and 2.1.330 suggests Anthropic is pushing toward more robust automation (longer task windows, safer destructive commands) and richer in-session visualisation (easel). If you are on a Pro or Max plan, checking /config to understand what easel does in your workflow is a reasonable first step. If you manage an enterprise deployment, the Admin Console sign-in integration and the self-hosted runner changes in the full changelog at code.claude.com deserve a careful read before your next update cycle.
