Anthropic announced Claude Code Routines on April 14, 2026, transforming Claude Code from an interactive coding assistant into an autonomous development automation platform. Routines are saved Claude Code configurations consisting of a prompt, repositories, and connectors that execute independently on Anthropic-managed cloud infrastructure—continuing to work even when a developer's laptop is closed.
Routines Enable Three Types of Automated Triggers
Claude Code Routines support three distinct trigger mechanisms:
- Scheduled triggers: Run on recurring cadences including hourly, daily, weekdays, or weekly intervals, with custom cron expressions supported via CLI with minimum 1-hour intervals
- API triggers: Each routine receives a dedicated HTTP endpoint that accepts POST requests with bearer token authentication, returning a session ID and URL
- GitHub triggers: Subscribe to repository events through the Claude GitHub App, including pull requests, pushes, issues, releases, PR reviews, comments, check runs, and workflow runs
Real-World Use Cases Span Development Lifecycle
Anthropic's documentation highlights six primary applications:
- Backlog maintenance: Nightly automated runs against issue trackers that apply labels, assign owners based on code areas, and post summaries to Slack
- Alert triage: Monitoring tools trigger routine APIs when error thresholds are crossed, with routines pulling stack traces, correlating with recent commits, and opening draft PRs with proposed fixes
- Bespoke code review: GitHub triggers on pull request creation apply team-specific review checklists and leave inline comments for security, performance, and style issues
- Deploy verification: CI/CD pipelines call routine APIs after production deploys to run smoke checks, scan error logs, and post go/no-go decisions to release channels
- Documentation drift detection: Weekly scans of merged PRs flag documentation referencing changed APIs and open update PRs
- Library porting: GitHub triggers on merged PRs in one SDK repository automatically port changes to parallel SDKs in other languages
Technical Architecture and Limitations
Routines run as full Claude Code cloud sessions with no permission-mode picker and no approval prompts. Each routine can combine multiple trigger types simultaneously and utilize all connected MCP connectors including Slack, Linear, and Google Drive. Daily run caps vary by plan: Pro receives 5 runs, Max receives 15, and Team/Enterprise receive 25. During the research preview, GitHub webhook events have per-routine and per-account hourly caps.
Anthropic's Rapid Product Expansion Strategy
The Routines launch represents Anthropic's third major release in six weeks, following Claude Code (the IDE layer) and Claude Managed Agents (orchestration layer). Industry observers characterized this pace as "not a product roadmap, it's a land grab" and noted that Anthropic is "not competing with Cursor anymore," signaling a shift toward comprehensive autonomous development infrastructure rather than traditional coding assistants.
Key Takeaways
- Claude Code Routines execute autonomously on Anthropic's cloud infrastructure, running independently of users' machines with scheduled, API, and GitHub event triggers
- Routines support real-world development workflows including automated code review, alert triage, deployment verification, and cross-repository library porting
- Daily run limits range from 5 (Pro) to 25 (Team/Enterprise), with routines drawing from subscription usage limits
- Anthropic shipped three major releases in six weeks: Claude Code, Claude Managed Agents, and Routines, representing a comprehensive automation stack
- The feature is available in research preview for Pro, Max, Team, and Enterprise plans with Claude Code enabled on the web