How does Anthropic's Claude Code routines work?

Anthropic adds repeatable “routines” to Claude Code

Anthropic has introduced a repeatable routines capability for Claude Code, currently described as a research preview. The feature is meant to let developers schedule and automate software development tasks through the coding agent, turning ad-hoc AI help into a more ongoing workflow.

What changed

Instead of running code-related requests only when prompted manually, routines allow developers to define repeatable actions and have them executed on a schedule. In other words, Claude Code can be set up to perform development tasks without requiring a fresh prompt every time.

Why it matters

That shift is important for teams that want AI assistance to behave more like an engineering tool—something that can: - run regularly (for example, periodic checks or scripted coding work), - reduce manual overhead, - and improve consistency by executing the same automation pattern repeatedly.

The news also reflects a broader trend in “agentic” tooling: vendors are trying to move beyond chat-style assistance and toward task orchestration that fits into established development practices.

What’s still unclear

The provided stories characterize the release as a research preview, but they don’t include implementation details such as how routines are configured, what triggers/actions are supported, or how Anthropic handles safety controls for automated software changes. Those specifics will likely determine how widely teams can adopt the capability.