Skip to content

Claude Code v2.1.330 adds enterprise sign-in and easel charts

The release brings organization-managed login, a built-in chart mod and new self-hosted runner options, alongside a long list of fixes.

Draftship Editorial3 min read

What shipped on 20 November

Claude Code enterprise sign-in is the first item in the changelog for v2.1.330, which Anthropic published on 20 November according to the timestamp on the project's GitHub releases page. That page shows a day and a time but no year. Alongside the sign-in change, the notes add easel, a built-in mod that draws charts, and several additions to self-hosted runner support. Most of the changelog is fixes, many of them aimed at long sessions and concurrent processes.

How the organization sign-in works

The source describes the sequence but not the internals. After claude auth login, Claude Code applies the organization's settings and uses its cloud provider or gateway. That is all the notes say: there is no list of supported cloud providers or gateways, and no description of which organization settings get applied.

There is a smaller login addition. Setting CLAUDE_CODE_FULL_SIGN_IN_LINK=1 prints the full sign-in link, for cases where the short link shows an error or a script needs to read values from it.

easel draws charts inside the session

easel is described as a built-in mod that draws charts when asked, and also under MCP metrics results. It runs on Pro, Max and, where available, Team and Enterprise, on macOS and Linux. Automatic charts can be turned off in /config.

The tool can also be caught by enterprise policy. One fix stops an "MCP server blocked by enterprise policy: easel" warning appearing at every start when an MCP allowlist or denylist excludes it. The tool stays blocked in that case; only the repeated warning goes away.

Self-hosted runner changes

Runners get most of the new surface area in this version:

  • --start-dir and SELF_HOSTED_RUNNER_START_DIR make Claude begin work in a folder of a repository, with settings and hooks loading as they would without it.
  • Two new /metrics values: claude_code_self_hosted_runner_order_to_registered_seconds (work order issued to registration accepted, on orchestrator-spawned runners) and claude_code_self_hosted_runner_order_to_first_session_seconds (work order to first session handed over, on runners started for a session).
  • A notice in cloud and self-hosted runner sessions when a plugin marketplace cannot be fetched because its repository is not one of the session's.
  • A notice when Anthropic-managed git refuses a plugin's own repository, naming the plugin and what to do.

Plugin installs also return more in JSON output: commandPrinted is true when the plugin's install command wrote to stderr, and commandFailed is set when it exited with an error, was killed or timed out.

Fixes aimed at long sessions

Several fixes target stalls and data loss rather than features. Claude Code no longer freezes for up to several minutes right after compacting a very long conversation, and WebFetch no longer hangs the session for minutes on some pages or returns empty results on others. Leaving a worktree no longer deletes one that another running Claude Code session is working in. Concurrent processes saving user settings no longer lose those changes, and /rewind no longer restores an older version of a file when two queued prompts run together. Turns no longer fail with "JSON schema is invalid", and session files no longer grow by many megabytes when many MCP tools carry an input schema the API refuses.

Limits and open questions

The notes do not say when the sign-in reaches all customers, which gateways are supported, or what a "mod" is beyond being built in. easel is listed for macOS and Linux only, with no mention of Windows. The --start-dir entry does not say how it interacts with path-scoped rules, and the two metrics names come without a stated retention or export format.

What it means for building apps with AI

Organization-managed sign-in matters if you run Claude Code inside a company: policy is applied at login and traffic goes through the organization's provider or gateway, which is the sort of thing a security review asks about. It governs access to the coding tool, not to the apps you build, and the two are easy to conflate. If you are adding sign-in to your own product, that is a separate job, and Draftship's docs on managing login and registration cover the app-side version.

Primary sources

  • v2.1.330 · Claude Code releases, Nov 20, 2026

This article was drafted with AI assistance from the primary sources listed below and checked automatically against them. An editor hasn't reviewed it yet.