ToolClaudeClaude Code

Anthropic adds security warnings to Claude Code

Original title: Claude Code 2.1.317

Plain language

What it is: Anthropic released a new update for Claude Code with new security warnings, cost reports for administrators, and bug fixes. The update improves user control over data sent to the network and scheduled tasks.
How it works: The system now displays a warning when a new cloud scheduled task gets unexpected connectors and explains how to remove them. It also asks for permission before web or shell tools send your account email address to the network.
Where it is available: The changes are available in Claude Code version v2.1.317, affecting Claude Desktop, the Claude apps gateway admin API, the VSCode extension, and cloud sessions.
Example: Administrators can use the new API endpoint to read past spend in total, per developer, per day, or per identity provider group.

What it is: Claude Code version v2.1.317 has been released, introducing security enhancements, telemetry updates, and extensive bug fixes. Key updates include blocking failed SDK callbacks, expanding OpenTelemetry metrics, and adding a gateway cost API.
How it works: Developers can now use onFailure: "block" for Agent SDK hook callbacks to prevent actions from running upon failure. Additionally, new metrics like anthropic.plan and anthropic.custom_base_url have been added to OTel spans and events.
Where it is available: The updates apply to the Claude Code CLI, Agent SDK, Claude apps gateway, and the VSCode extension.
Example: A developer running a web or shell tool will now be prompted to approve the action before their account email address is transmitted over the network.

The headline and explanation on this page are based on Anthropic's post. The full details are in the original.

Claude Code 2.1.317