Routines
person concept tool org talk claim — click a node to jump to its page; hover an arrow for the relation
Routines is a feature of Claude Code that enables proactive, scheduled, and event-driven agent workflows by abstracting away infrastructure concerns such as hosting, session state, and external connectors. Described as "brand new" at launch, they represent what Boris Cherny calls "the first really obvious application" of programmatic Claude Code use via the agent SDK.
Core Purpose: Replacing Manual Infrastructure
The central motivation behind Routines is eliminating the need for developers to manage their own scheduling and automation infrastructure. Maya, an advocate for the feature, frames it explicitly as an alternative to Cron jobs and similar tooling: "we built routines so you don't have to focus on maintaining all of this infra". Concretely, the feature handles "the hosting, the session state, and the connector all for you", allowing engineers to focus on agent logic rather than plumbing.
Integration with External Services
Routines are designed to interact natively with the tools developers already use. They support GitHub event triggers — "we have the ability to work natively with GitHub events" — and can push notifications through Slack: "I'll actually want it to ping me on Slack". A concrete example is the Deploy Verifier routine, which monitors service health after a deployment: "this deploy verifier... you want to make sure that this service is healthy".
Agent-on-Agent Patterns
Routines enable compositional, multi-agent workflows. One highlighted pattern is the generator-critiquer (or reviewer) setup, where one routine generates outputs and a second reviews them: "You can actually set up one routine to go ahead and create docs PRs, and you can set up another routine" to critique or verify those outputs. This agent-on-agent review pattern is presented as a method for improving quality in automated pipelines.
Context as the Limiting Factor
A key design principle governing Routines is that the quality of context provided to Claude determines the ceiling of what any routine can accomplish: "whatever context Claude has, that's the ceiling of how successful Claude will be". This frames context engineering as the primary lever for improving routine performance, not the scheduling or infrastructure layer.
Advocacy
Both Maya and Boris Cherny are enthusiastic proponents. Maya emphasizes the infrastructure abstraction angle, while Cherny positions Routines as the natural entry point for developers beginning to use Claude Code programmatically: "routines are the first really obvious application". Cherny also singles them out as a highlight when reflecting on a year of Claude Code: "One that I'm super excited about is routines."