Descriptions:
Anthropic’s Applied AI team member Maya introduces Routines, a brand-new feature inside Claude Code designed to transform the tool from a reactive coding assistant into a proactive coding teammate. The session, delivered at Anthropic’s first ‘Code with Claude’ developer event, addresses three core pain points developers face when building proactive agents today: choosing where agents run, deciding when they should trigger, and maintaining visibility into agent sessions in real time.
Routines solves these challenges by running Claude Code on Anthropic’s managed infrastructure, eliminating the need for self-hosted cron jobs or custom endpoints. Developers define just four things — a prompt, connected repositories, available connectors (such as Google Drive or Slack), and a trigger — and Claude handles the rest. Triggers can be time-based schedules or event-based, with native support for GitHub events like pull request merges and custom webhooks.
The talk walks through a real internal use case at Anthropic: automating documentation creation by diffing the Claude Code source repo against the docs repo whenever a release is cut. Maya covers practical design considerations including context provisioning, steerability strategies, and the generator-critique pattern for agent-on-agent quality review. This session is essential viewing for developers building autonomous workflows on top of Claude Code.
📺 Source: Claude · Published May 20, 2026
🏷️ Format: Keynote Launch
