Products

Anthropic Launches Claude Code Routines for Scheduled, Cloud-Hosted Autonomous Agent Workflows

On April 14, 2026, Anthropic released Claude Code Routines in research preview: users define a prompt, connect GitHub repositories and external tools, and set triggers (cron schedules, API calls, or GitHub events) for the agent to run autonomously on Anthropic-managed cloud infrastructure without requiring a local machine to stay online.

Tech Insights Reporter Apr 14, 2026 7 min read San Francisco

title: "Anthropic Launches Claude Code Routines for Scheduled, Cloud-Hosted Autonomous Agent Workflows" summary: "On April 14, 2026, Anthropic released Claude Code Routines in research preview: users define a prompt, connect GitHub repositories and external tools, and set triggers (cron schedules, API calls, or GitHub events) for the agent to run autonomously on Anthropic-managed cloud infrastructure without requiring a local machine to stay online." category: "Products" author: "Tech Insights Reporter" date: "2026-04-14" readTime: "7 min read" location: "San Francisco" hue: 220 regions: - "us" platforms: - "anthropic" tag: "Developer Tools" featured: false

TLDR

Anthropic launched Claude Code Routines on April 14, 2026, as a research preview feature for Pro, Max, Team, and Enterprise plans. A Routine packages a saved Claude Code configuration — including a prompt, one or more GitHub repositories, and connectors — that executes automatically on Anthropic’s cloud. Triggers include cron-style schedules, incoming API/webhook calls, or GitHub events such as pull_request.opened or issues. The launch coincided with a major redesign of the Claude Code desktop and web experience adding multi-session sidebar management, an integrated terminal, parallel agent support, and a faster diff viewer. Routines enable “infra-less” agentic automation: the AI wakes on trigger, performs work, and sleeps without local compute or always-on servers.

What Claude Code Routines Provide

Routines turn Claude Code from an interactive, session-based coding agent into a scheduled and event-driven automation platform:

Configuration: Define a prompt describing the task, point it at GitHub repo(s), and attach connectors or tools.
Triggers:
- Recurring schedules (cron-like).
- API calls (POST to a /fire endpoint with beta header experimental-cc-routine-2026-04-01).
- GitHub events (e.g., PR opened, issue created, webhook).
Execution: Runs entirely on Anthropic-managed infrastructure. No requirement for the user’s Mac, PC, or self-hosted server to be powered on or connected.
Access: Configured via claude.ai/code/routines, the desktop app, or CLI (/schedule). Available in research preview with plan-based daily run caps.

The feature ships alongside broader Claude Code improvements for handling longer sessions and parallel work.

How It Works in Practice

Users create a Routine once. On trigger:

The agent wakes in the cloud.
It receives the prompt and context (repo state, prior state if maintained).
Claude Code reads files, runs commands, edits code, and uses connected tools.
Results (diffs, logs, outputs) are delivered via the interface or notifications.
The session ends; the agent returns to dormant state until the next trigger.

This supports unattended workflows such as:

Nightly code reviews or labeling of new issues.
Automated responses to Sentry alerts or monitoring webhooks.
Scheduled refactoring, test generation, or dependency updates.
CI/CD-adjacent tasks triggered by GitHub activity.

Early coverage and user reports highlight the shift from “keep my laptop on” or custom server setups to fully managed cloud execution.

Research Preview Limitations and Access

As a research preview:

Rate limits and quotas apply (plan-dependent daily runs).
APIs and behaviors may evolve (beta header required for triggering).
Best suited for well-scoped, repeatable tasks rather than fully open-ended work.
Users must still review outputs for correctness and security, consistent with other agentic tools.

Anthropic positions it as an early step toward reliable, always-available agentic coding infrastructure.

Why this story matters

Claude Code Routines mark a practical expansion of agentic AI from interactive assistance to background automation. By moving execution to managed cloud with native support for schedules and real developer events (GitHub webhooks, API), Anthropic removes significant DevOps and uptime friction for teams wanting AI to handle routine or time-based coding work. The April 14, 2026 preview, paired with app redesigns for parallelism and longer sessions, signals the maturation of coding agents into production-adjacent tools. For developers and engineering organizations, it lowers the barrier to reliable AI-driven workflows while raising new questions around monitoring, cost control, review processes, and the scope of tasks safely delegated to autonomous agents.

Sources

Anthropic Claude Code documentation on Routines (launch context and API details, April 2026). https://docs.anthropic.com/en/docs/claude-code/routines
9to5Mac: “Anthropic adds routines to redesigned Claude Code, here’s how it works” (April 14, 2026). https://9to5mac.com/2026/04/14/anthropic-adds-repeatable-routines-feature-to-claude-code-heres-how-it-works/
Multiple contemporaneous reports and user guides (Medium, Substack, DevOps.com) confirming April 14 research preview launch, trigger types, cloud execution model, and integration with the Claude Code redesign.
Anthropic research and engineering posts on Claude Code usage patterns around the same period.

Featured Image Alt Text

Illustration of Claude Code Routines running on cloud servers: a sleeping laptop on a desk beside an active cloud icon with scheduled cron, GitHub PR, and API webhook triggers firing autonomous Claude agents that edit code repositories, dated April 14 2026 research preview