Claude Code Routines: Scheduled, API, and GitHub-Triggered Automations

Claude Code Routines: Cloud Automation Without a Local Machine

Anthropics Claude Code now supports automated, unattended execution through a new feature called Routines, launched in research preview on April 14, 2026. Routines allow developers to configure Claude Code sessions to run on a defined schedule, in response to an HTTP API call, or triggered by GitHub events — all without requiring the developer's computer to be online or running.

The feature represents a significant expansion of Claude Code's scope from an interactive coding assistant to an autonomous automation platform.

Three Trigger Types

Routines support three distinct ways to initiate a Claude Code session automatically:

Scheduled Runs

Developers can configure a Routine to run on a cron-style schedule — for example, every night at 2 AM to triage new GitHub issues, scan for security vulnerabilities, or generate a weekly code quality report. Scheduled Routines run on Anthropic's cloud infrastructure, requiring no local process to be running.

HTTP API Triggers

A Routine can be exposed as an HTTP endpoint and triggered programmatically from any external system. Developers authenticate requests using bearer tokens and call the endpoint with a payload that describes the task. This makes it straightforward to integrate Claude Code into CI/CD pipelines, internal tooling, or third-party automation workflows.

GitHub Webhook Triggers

Routines can subscribe to GitHub webhook events — such as a pull request being opened, a commit being pushed, or an issue being labeled. When the specified event fires, Claude Code automatically runs the configured session. This enables use cases like automated code review on every PR, documentation updates triggered by code changes, or test generation on new commits.

Cloud Execution

All Routines run on Anthropic's cloud infrastructure rather than on the developer's local machine. This means the developer's computer does not need to be powered on or connected to the internet for a Routine to execute. Sessions run in isolated environments with access to the configured repository, secrets, and tool permissions.

Cloud execution also means Routines are resilient to the transient nature of developer laptops — they fire reliably on schedule regardless of whether the developer is at their desk.

Use Cases

Anthropics documentation for Routines highlights several representative use cases:

• Nightly bug triage — Claude Code reviews new GitHub issues each night, labels them, assigns priority, and drafts initial responses
• Docs drift detection — Claude Code checks whether documentation is still accurate relative to recent code changes and opens a PR with corrections
• Automated code review — every pull request triggers a Claude Code session that reviews the diff, checks for common issues, and posts a review comment
• Dependency monitoring — a weekly Routine scans for outdated or vulnerable dependencies and opens PRs to update them

Plan Limits and Research Preview Status

Routines are available in research preview, meaning the feature is functional but subject to change based on feedback. Daily execution limits vary by plan:

• Pro: 5 runs per day
• Max: 15 runs per day
• Team and Enterprise: 25 runs per day

Anthropics research preview designation signals that the team is actively gathering developer feedback on trigger types, execution limits, and use case fit before committing to a final API surface.

Authentication and Security

API-triggered Routines use bearer token authentication. Developers generate tokens from their Claude Code settings and include them in the Authorization header of incoming API calls. GitHub webhook triggers use webhook secrets for payload verification, following the standard GitHub webhook security model.