Claude Code launches Routines: scheduled jobs, API triggers, and event-driven workflows—AI automation no longer needs to run with your computer turned on

Anthropic’s Claude Code, an AI development tool under the Anthropic umbrella, has officially launched a new “Routines” feature, enabling developers to trigger AI Agents to automatically run tasks via scheduling, API calls, or Webhook events—and it requires no need to keep the computer powered on. This feature is currently available in the form of a Research Preview, and all users on paid Claude Code plans can use it on the web. (Source)

What are Routines?

Routines are Claude Code’s new automation framework, upgrading the /schedule command that used to be limited to the CLI into a complete cloud-based automation system. If users previously set up scheduled tasks, those tasks will be automatically converted into Routines, with no manual migration needed.

Routines run on Anthropic’s web-based infrastructure, which means that even if you close your laptop or leave the office, the AI Agent will still automatically execute according to the configured conditions. Each Routine gets its own dedicated API endpoint, making it easy to integrate into existing development workflows.

Three types of triggers

Routines provides three different triggering mechanisms, covering most automation scenarios:

Trigger type | Description | Use cases
Scheduling (Cron) | Run on a schedule based on a cron expression | Every day at 2 a.m., fetch the bug list from Linear and fix the API endpoint (POST)
Through HTTP POST requests | Triggered from a monitoring alert system | From an oncall classification Webhook, listen for external events (e.g., GitHub) | Monitor whether PRs involve changes to the authentication module

Practical use cases

The goal of Routines’ design is to help AI Agents fit into a team’s day-to-day development workflow. Here are a few typical usage scenarios:

In scheduling scenarios, developers can configure Claude Code to automatically scan the bug list on Linear every day at 2 a.m., attempt fixes, and submit a PR. In Webhook scenarios, when a new Pull Request appears on GitHub, a Routine can automatically check whether it involves code changes related to identity verification, and immediately notify the team if any issues are found. API triggers are well-suited for integrating with existing alerting systems—when PagerDuty or other monitoring platforms emit an alert, the AI Agent is automatically started to perform initial event classification and diagnosis.

Integration with existing features

Routines can be seen as an extension of Claude Code’s dynamic loops and event-driven architecture. In the past, Claude Code has already demonstrated capabilities in code generation, debugging, and refactoring, and Routines allow these capabilities to run automatically without human oversight.

For teams evaluating AI development tools, Routines provides an upgrade path from “manually triggering AI-assisted work” to “AI executing automatically.” To learn more about Claude’s features and how to use them, you can refer to the complete Claude guide; for readers interested in the technical architecture of AI Agent frameworks, you can also read related analysis.

This article, “Claude Code Launches Routines: Scheduling, API Triggering, Event-Driven—AI Automation No Longer Needs to Keep Your Computer On,” first appeared on Chain News ABMedia.