The landscape of artificial intelligence agents has just taken a massive leap forward. In mid-April 2026, Anthropic unveiled its latest major innovation in a research preview: Claude Routines.
While Claude Code had already established itself as a powerful local "sidekick" for developers, Routines transforms the tool into a full-fledged autonomous orchestration platform operating directly on Anthropic’s managed cloud infrastructure.
What Exactly is Claude Routines?
At its core, Claude Routines enables technical teams to package a prompt, a code repository, and various connectors into a "routine". This routine executes entirely autonomously, without requiring a human developer to approve every step or validate execution locally.
The most disruptive aspect is its "Always-On" nature: because it runs on Anthropic’s servers, these routines keep working even if you close your laptop or are away from the office.
Three Ways to Trigger Automation
Anthropic has designed an extremely versatile trigger system, perfectly suited to integrate with existing professional workflows:
- Schedules (Cron): You can set up recurring tasks, such as "review and refactor legacy code every Friday at 6:00 PM," or "summarize the server error log first thing every morning."
- API Calls (HTTP POST): This allows hooking Claude Routines into any external infrastructure, such as a CRM, an ERP, or flows built in Zapier/Make.
- Events (GitHub Integration): This is the crown jewel for development teams. Routines can be triggered automatically when a Pull Request is opened, when code is pushed, or when an Issue is created. This allows the AI to perform security audits, deep code reviews, or automated bug fixing before a human ever intervenes.
Autonomous Execution and Interface Redesign
Until now, using advanced models in the terminal often required the user to actively monitor operations, approve permissions, or guide the model. Routines eliminates this bottleneck. By packaging the workflow as a routine, you grant Claude Code the agency to act independently throughout that specific flow, drastically accelerating the pace of work.
Alongside this release, Anthropic has completely redesigned the Claude Code desktop interface. It now features a sidebar to manage multiple active sessions side-by-side, an integrated terminal, a native in-app file editor, a blazingly fast diff viewer, and HTML/PDF preview capabilities. The entire UI is now a fully customizable drag-and-drop workspace.
Usage Limits by Plan
Given the immense computational resources consumed by fully autonomous cloud agents, Anthropic has established initial usage limits:
- Pro Plan: 5 routines per day.
- Max Plan: 15 routines per day.
- Team / Enterprise Plans: Up to 25 routines per day.
Why is this Crucial for your SME?
For small businesses and development agencies, the adoption of Claude Routines means you no longer need to allocate your most talented (and expensive) engineers to routine maintenance tasks, such as minor refactoring, PR reviews, or responding to known Tier 1 incidents.
You are effectively hiring an untiring "DevSecOps Engineer" that works 24/7 in the cloud, operating directly on your repositories and acting only when your business logic dictates. The future of software is not just writing code with AI; it's about autonomous operations and maintenance.