Anthropic Launches Claude Routines: Automated Task Execution on Schedule, API Calls, and Events
Key Takeaways
- ▸Routines enable Claude to execute automated tasks on schedules, via API calls, or in response to events, eliminating the need for users to maintain continuous connectivity
- ▸Each routine receives a dedicated API endpoint, enabling seamless integration with alerts, deployment hooks, and internal tools for production workflows
- ▸The feature supports three execution modes: scheduled tasks (e.g., nightly bug fixes), webhook-driven event responses (e.g., GitHub PR monitoring), and on-demand API-triggered automation
Summary
Anthropic has introduced Routines in Claude Code, a new research preview feature that enables users to configure automated tasks that run on Anthropic's web infrastructure without requiring users to keep their devices powered on. Routines can be triggered through multiple mechanisms: scheduled intervals, API calls, or event-based triggers, and each routine receives its own dedicated API endpoint for integration with external tools and services. This capability allows developers to automate workflows such as monitoring GitHub repositories, responding to alerts, and performing scheduled maintenance tasks directly through Claude.
The feature offers three primary use cases: API-triggered routines for responding to alerts and webhook payloads, event-based routines that subscribe to GitHub events and respond in real-time (with more event sources coming soon), and scheduled routines that execute on defined cadences—such as pulling top bugs from Linear at specific times and automatically generating draft pull requests. The Routines feature is now available across all paid Claude plans for users with Claude Code enabled on the web, with no migration required for users who previously utilized the /schedule command in the CLI.
- Routines are now available across all paid plans with Claude Code on the web, with expanded event source support planned for future releases
Editorial Opinion
Claude Routines represent a significant step toward making AI agents practical for real-world software development workflows by bridging the gap between interactive AI assistance and autonomous background task execution. The ability to trigger Claude through multiple channels—schedules, webhooks, and APIs—and have it operate on cloud infrastructure creates genuine opportunities for automating DevOps and engineering tasks without custom infrastructure. However, the success of this feature will depend on how reliably Claude handles edge cases in production environments and whether the feature set expands quickly to support diverse event sources and integrations.
