Anthropic is expanding the capabilities of its Claude Code tool with the release of 'routines,' a new feature currently in research preview. The feature allows developers to package prompts, repositories, and connectors into automated configurations that run on Anthropic-managed cloud infrastructure.
According to documentation from code.claude.com, routines are designed to execute tasks even when a user's local machine is powered down. The system supports several trigger types, including scheduled cadences like nightly or weekly runs, manual triggers via HTTP POST requests to an API endpoint, and GitHub-based automation responding to pull requests or pushes.
Automated developer workflows
The documentation outlines several practical applications for the new automation layer. For instance, a routine can be configured to handle backlog maintenance by running every weeknight to read issues from a tracker, apply labels, and post summaries to Slack.
Developers can also use the API trigger for alert triage. In this scenario, a monitoring tool calls the routine when an error threshold is met, prompting the AI to pull stack traces, correlate them with recent commits, and open a draft pull request with a proposed fix.
For teams focused on code quality, a GitHub trigger can run automatically when a pull request is opened. The routine can then apply specific checklists for security, performance, and style, leaving inline comments to assist human reviewers.
Anthropic notes that routines are available for Pro, Max, Team, and Enterprise plans, provided Claude Code on the web is enabled. Users can manage these automated tasks through the Claude.ai interface or via the command-line interface using the /schedule command.
Because the feature is in research preview, the documentation warns that 'behavior, limits, and the API surface may change' as the technology matures.
