Anthropic Launches Claude Code Routines for Automated Development Workflows

Anthropic has rolled out a research preview of routines for Claude Code, allowing developers to configure AI-powered automations that run on schedules, respond to API calls, or trigger from GitHub events—all without keeping a laptop open or managing custom infrastructure.

The feature addresses a friction point that’s plagued developer automation: managing cron jobs, servers, and tooling just to get an AI assistant to handle repetitive tasks. Routines bundle prompts, repository access, and connectors into packages that execute on Anthropic’s web infrastructure.

Three Flavors of Automation

Scheduled routines work on hourly, nightly, or weekly cadences. A team could configure one to pull the top bug from Linear at 2am, attempt a fix, and open a draft PR before anyone wakes up. Developers already using the /schedule command in Claude Code’s CLI will find those tasks automatically converted to scheduled routines.

API routines get their own endpoints and auth tokens. POST a message, get back a session URL. This opens integration possibilities with alerting systems, deployment pipelines, and internal tools. One use case already emerging: Datadog alerts trigger a routine that pulls traces, correlates them with recent deployments, and drafts a fix before on-call engineers even open the page.

Webhook routines—starting with GitHub—subscribe to repository events. Claude spins up a session for each matching PR and continues monitoring for comments and CI failures. Teams are using this for automated code review against custom checklists and even cross-language library ports, where a Python SDK change automatically generates a matching Go SDK PR.

Early Adoption Patterns

Beta users have gravitated toward a few workflows. Nightly backlog triage labels and assigns new issues, then posts summaries to Slack. Weekly docs-drift scans catch documentation that references changed APIs and open update PRs automatically. Deploy verification routines run smoke checks and scan error logs, posting go/no-go decisions to release channels.

The feedback-to-fix pipeline stands out: a docs feedback widget sends reports directly to a routine, which opens a session with the relevant repo and drafts changes without human intervention on the initial pass.

Access and Limits

Routines are live now for Pro, Max, Team, and Enterprise subscribers with Claude Code on the web enabled. The daily caps: Pro gets 5 routine runs, Max gets 15, and Team/Enterprise users get 25. Additional runs consume extra usage allocation. All routine execution draws from standard subscription limits, same as interactive sessions.

Anthropic says webhook triggers will expand beyond GitHub to additional event sources, though no timeline was provided. For developers tired of maintaining automation infrastructure, the pitch is straightforward: define the task once, let it run in the cloud.

Image source: Shutterstock