Anthropic Adds Scheduled Routines to a Redesigned Claude Code

Anthropic shipped two meaningful upgrades to Claude Code on April 14: a routines system that runs automations on remote infrastructure, and a redesigned interface that lets developers manage multiple sessions from a single window. Neither change is cosmetic. Both reflect a deliberate push to position Claude Code as a development platform rather than an enhanced coding assistant.

The routines feature is the more consequential addition. Previously, users who wanted recurring automations had to wire together cron jobs, external infrastructure, and tools like MCP servers on their own. According to 9to5Mac, routines bundle that plumbing directly into the product. A developer defines a task, sets a schedule or trigger, and the system executes it on Claude Code's web infrastructure. The user's machine does not need to be online.

Each routine arrives pre-connected to the user's repositories and configured connectors, cutting setup overhead compared to assembling custom toolchains. Documented use cases include scheduled code tasks, API-driven workflows, and GitHub integrations. The feature launches today as a research preview for Pro, Max, Team, and Enterprise subscribers.

The infrastructure shift

The architectural move here is significant. Anthropic is absorbing the execution layer that developers previously had to manage themselves, which puts Claude Code closer to what CI/CD platforms offer: persistent, scheduled, event-driven compute with an agent that understands code context. That is a different product category than a chat interface or autocomplete tool.

According to The Tech Portal, Anthropic is also preparing its next flagship model, tentatively called Opus 4.7, with a stated focus on multi-step reasoning, longer-horizon task handling, and coordination between multiple AI agents. Routines look like early infrastructure for that roadmap: a way to establish scheduling and execution primitives before the underlying models are capable enough to make fully autonomous operation reliable.

Alongside the scheduling system, Anthropic released a redesigned Claude Code interface. The updated layout lets users run multiple Claude sessions in parallel from a single window, managed through a new sidebar. The environment now includes an integrated terminal, file editor, HTML and PDF preview pane, and a faster diff viewer, all configurable via drag and drop. As 9to5Mac notes, Claude Code, Cowork, and Chat now all live within the same application, a consolidation that echoes how traditional IDEs gradually absorbed adjacent developer workflows.

What it means for practitioners

For engineers who already use Claude Code daily, the interface changes are immediately practical. Multi-session support removes a common friction point: iterating on one task while running tests in another no longer requires managing separate windows. The integrated terminal and file editor reduce context-switching mid-session, and the diff viewer improvement, modest in isolation, matters because code review is a high-frequency workflow where small efficiency gains compound.

The routines system will take longer to assess. Research preview status means the feature is still being shaped by early feedback, and limiting access to paid tiers signals that Anthropic is watching capacity and quality signals before broadening rollout. The real test is failure handling: scheduled automations are only useful if they surface errors in actionable ways. That is an observability and reliability challenge that model intelligence alone cannot solve, and it is one Anthropic will need to address before routines graduate from preview.

Anthropologic introduced auto mode for Claude Code just last month, a release also covered by 9to5Mac, and the pace since then, auto mode, routines, a full interface redesign, all within weeks, suggests the team is moving quickly. Whether that pace produces a platform developers can depend on, or an accumulation of features still in preview, is the question routines will have to answer.

If reliability holds, the logical endpoint is a version of Claude Code that continues working while engineers are offline. That is not a hypothetical: it is exactly what scheduled server-side infrastructure is designed to enable. The open question is whether developer trust in autonomous code agents has caught up to the tools' ambitions.

FAQ

What are Claude Code routines?
Routines are scheduled or triggered automations in Claude Code that run on Anthropic's web infrastructure. Developers package tasks with access to their repos and connectors, set a schedule or trigger condition, and the system executes without requiring the local machine to be active.

Do I need to keep my computer online for Claude Code routines to run?
No. Routines execute on Claude Code's remote infrastructure, so they continue running even when your Mac is offline or asleep.

Which Claude plans include access to routines?
The feature is currently available to Pro, Max, Team, and Enterprise subscribers as a research preview, with tier-specific usage limits.

What changed in the Claude Code interface redesign?
The redesign adds multi-session support in a single window with a new sidebar, an integrated terminal, file editor, HTML and PDF preview, and a faster diff viewer. The layout is drag-and-drop configurable. Claude Code, Cowork, and Chat are now unified within the same Claude application.