On 20 November 2026 Anthropic released Claude Code 2.1.330. Per the Claude Code changelog, three changes stand out for anyone commissioning software built with AI coding tools.
Longer-running automation. Scheduled tasks and /loop now run for up to 30 days instead of 7, for as long as their session keeps running. Recurring jobs such as daily checks or reports no longer need restarting each week.
More guardrails on deletion. More rm (file deletion) commands aimed at the root, the home folder, a top-level folder or the working folder now ask for approval, or are refused where nobody can answer. That matters most for unattended sessions.
Enterprise sign-in. After claude auth login, Claude Code can sign in to organisations managed in Anthropic’s Enterprise Admin Console. It then applies the organisation’s settings and uses its cloud provider or gateway.
Also in the release
- A built-in mod called Easel draws charts on request and under MCP metrics results. Automatic charts can be switched off in
/config. deniedMcpServersname entries now hide only the named claude.ai connector. To block every connector at a URL, deny its exact URL.- VS Code gains
/forkand/workflows. - Fixes include freezes after compacting very long conversations, and
ExitWorktreeno longer deleting a worktree another session is using.
Why it matters
Two themes repeat across recent releases: AI tools are being left to run longer on their own, and vendors are adding safeguards for that. A tool that works unattended for a month needs clear limits on what it can delete or change, and organisation-level control over who signs in and where the work runs.
So what
If a supplier uses AI coding tools on your product, ask what runs unattended, what it is allowed to change, and who approves destructive actions. Our AI-assisted development approach keeps a named engineer accountable for every change. For a product built to your needs, see custom software development, or start a project.