Anthropic is taking a new step in the development of AI-powered software development with the introduction of so-called routines in Claude Code. The feature, currently in a research phase, helps developers automate recurring tasks within their workflows without requiring additional infrastructure.
With routines, users can define a task once—including access to codebases and external links—after which it is automatically executed based on a schedule, an API call, or an event. This shifts part of the management that previously fell to developers themselves to Anthropic’s infrastructure. Tasks that previously depended on local scripts or external tools now run entirely within Claude Code’s cloud environment and remain active even if a developer shuts down their system.
It is striking how Anthropic positions the feature relative to existing solutions. Routines share similarities with cron jobs and GitHub Actions, but do not execute fixed scripts. Instead, an AI model controls execution based on context. At the same time, they remain more limited than full-fledged AI agents, as they are not continuously active and do not maintain a persistent state, according to SiliconANGLE, making routines best viewed as a hybrid between traditional automation and agent-based AI.
The feature responds to a broader trend in which AI not only supports individual programming tasks but also takes over processes. Developers are already using such systems for code generation and analysis, but routines enable automated workflows, such as checking bug lists or responding to reports after a software rollout.
Routines respond to code changes and API calls
Anthropic is focusing on integration with existing development environments. Routines can respond to events in GitHub repositories and automatically start a session for each change, which remains active as long as updates continue. External systems can trigger a routine via an API, after which Claude Code independently performs an analysis and returns a result.
According to Anthropic, teams are experimenting with applications such as automatically triaging issues, updating documentation, and performing deployment checks. In some cases, suggestions for code changes are also generated, though human review remains necessary.
In addition to routines, Anthropic is working on the Claude Code user interface. The desktop application has been redesigned to directly integrate features such as a terminal, diff viewer, and editor. With this, the company aims to make developers less dependent on external tools, according to SiliconANGLE, which sees this as part of a broader strategy to build a central development interface around the AI tool.
Claude Code now also supports multiple concurrent sessions, which aligns with how developers are increasingly using AI: performing multiple tasks in parallel and making adjustments along the way.
Routines are available to users of the paid versions of Claude Code, with limits on the number of daily executions. Additional usage is billed separately, while Anthropic investigates how this form of automation is applied in practice.