Claude Code Routines: The Desktop Redesign That Changes How Developers Work

If you've been using Claude Code as your AI coding assistant, April 2026 just changed the game. On April 14, Anthropic shipped a major desktop redesign that goes far beyond a visual refresh — the headline feature is Routines, a way to run automated coding tasks in the cloud so your laptop doesn't need to stay on.

This is the kind of update that sounds small on paper but fundamentally reshapes your workflow. Here's what changed, how Routines work, and what each plan tier actually gets.

What Changed in the Claude Code Desktop Redesign

The redesign isn't just cosmetic. Anthropic rebuilt several core surfaces of the Claude Code desktop app simultaneously:

Multi-session sidebar. You can now manage multiple projects without switching contexts or losing your place. Each project lives in a dedicated sidebar panel — click between them the way you switch browser tabs. Rebuilt diff viewer. The previous diff view was functional but cramped. The new one gives you more screen real estate and a cleaner presentation of what changed, making code review inside Claude Code actually pleasant. Expanded preview area. Web and component previews now use more of the screen, which matters when you're iterating on UI components and need to see the full output without squinting. Integrated terminal and in-app file editor. Previously, Claude Code often required you to jump between the app and your system terminal or external editor. The new version brings both in-app, reducing context switching.

These are quality-of-life improvements, but they're not what the developer community is talking about. That would be Routines.

Claude Code Routines Explained

A Routine is a saved, automated task that Claude Code runs in the cloud — without you being present and without your laptop staying powered on.

Think of it like this: you define a task ("run my test suite and summarize failures", "check for security vulnerabilities in new PRs", "generate a changelog from today's commits"), save it as a Routine, and then schedule or trigger it. Claude Code executes it remotely and delivers the results when it's done.

This closes a significant gap in how AI coding assistants work today. Most tools — including Claude Code before this update — require an active session. If you close your laptop, the task stops. Routines offload execution to Anthropic's infrastructure, which means:

Long-running analysis tasks don't tie up your machine
You can kick off a Routine before a meeting and have results waiting when you're done
CI/CD-adjacent tasks (code review, test summarization) can run on a schedule without custom infrastructure

What can you automate with Routines?

The use cases that make the most sense right now:

Daily code review summaries — Run a Routine each morning that scans overnight commits and surfaces potential issues

Dependency audit reports — Schedule weekly checks for outdated packages or known vulnerabilities

Test failure summaries — After a CI run, trigger a Routine to explain which tests failed and why

Documentation drafts — Point a Routine at a module and have it draft JSDoc or API docs async

Codebase health reports — Periodic complexity and coverage reports without blocking your workday

The pattern is: anything you'd normally ask Claude Code to do in a session that (a) takes more than a few minutes or (b) you want to happen on a schedule is a Routine candidate.

Routines by Plan Tier: Who Gets What

Anthropic is rolling out Routines with per-day limits tied to your subscription tier:

Plan	Routines per Day
Pro	5
Max	15
Team	25
Enterprise	25

For most individual developers, the Pro limit of 5 routines/day is the right starting point. If you're primarily using Routines for morning code reviews, a dependency check, and the occasional test summary, 5 is plenty.

Max users (the $100/month plan that replaced Claude Pro + API credits) get 15 — enough for teams of 1-3 who want more automation throughput or are actively experimenting with different Routine configurations. Team and Enterprise users at 25/day are positioned for actual workflow integration: think multiple Routines per project, per day, coordinated across teammates.

One thing worth noting: Routines run against your existing Claude Code context and permissions. They're not a separate product — they're baked into the same session model, just executed asynchronously in the cloud.

How to Set Up Your First Claude Code Routine

Getting started with Routines in the new desktop is straightforward. Here's the general flow:

Step 1: Define the task clearly. Routines work best when the prompt is specific. "Review my code" is too vague. "Check src/api/ for any endpoints that don't have input validation and list them with line numbers" is the right level of specificity. Step 2: Test it as a regular session first. Before saving as a Routine, run the task in a normal Claude Code session to validate the output quality. This ensures your Routine prompt is producing what you expect. Step 3: Save as a Routine. In the redesigned interface, there's a dedicated "Save as Routine" action once you have a working session. Give it a descriptive name — you'll thank yourself when you have five of them. Step 4: Set the trigger. Routines support both scheduled (time-based) and manual triggers. For daily reports, use a schedule. For event-driven tasks (like post-CI summaries), use manual triggers or connect them via the Claude Code API. Step 5: Review outputs in the sidebar. Results appear in the new multi-session sidebar, clearly labeled by Routine name and timestamp. You can respond, iterate, or export from there.

What This Means for the Claude Ecosystem

The Routines launch is significant beyond the individual feature because it signals Anthropic's direction: Claude Code is becoming an agentic platform, not just an interactive assistant.

The sequence of moves over the last few months makes this clear:

February 2026: Claude Opus 4.6 and Sonnet 4.6 launch with extended context and improved instruction-following
March 2026: MCP (Model Context Protocol) hits mainstream adoption — Claude Code can now connect to databases, APIs, and external services through standardized connectors
April 8, 2026: Claude Managed Agents enters public beta — developers can deploy autonomous agent loops via REST API without building custom infrastructure
April 14, 2026: Claude Code Routines launch — the same agentic capability comes to the desktop in a no-code form

The pattern: Anthropic is building a full stack for AI automation. The API layer (Managed Agents), the protocol layer (MCP), and now the desktop layer (Routines) are all pointing at the same destination — AI systems that execute real work autonomously, not just answer questions.

For developers preparing for the Claude Certified Architect (CCA) certification, this is directly relevant content. The CCA exam tests your understanding of agentic patterns, tool use, and multi-step execution — exactly what Routines represent in practice.

The New Features Worth Testing First

If you're upgrading to the redesigned desktop, here's a prioritized list of what to explore:

High value, low learning curve:

Multi-session sidebar — immediately useful if you work on more than one project
New diff viewer — better than before, no setup required

High value, worth investing time:

Routines — start with one daily report task and iterate from there
In-app terminal — reduces the need to context-switch out of Claude Code

Situational:

Expanded preview — essential for frontend work, less relevant for backend-focused teams

Key Takeaways

Claude Code's April 2026 desktop redesign introduces Routines — cloud-based automated tasks that run without your laptop
Routines limits: Pro = 5/day, Max = 15/day, Team/Enterprise = 25/day
The redesign also ships a multi-session sidebar, rebuilt diff viewer, expanded preview area, and integrated terminal
Routines represent Anthropic's broader push to make Claude Code an agentic platform, not just an interactive assistant
Best first Routine: a daily morning summary of new commits or test failures in your active project

Start Building With Claude's Full Stack

Claude Code Routines are the desktop face of a larger agentic architecture that now spans the Claude API (Managed Agents), the protocol layer (MCP), and the developer desktop. Understanding how these pieces fit together is increasingly important for anyone building or architecting AI systems.

If you're working toward the Claude Certified Architect certification, these are exactly the concepts you'll be tested on — agentic patterns, tool use, model selection, and system design with Claude at the center.

Practice CCA exam questions on AI for Anything →

You can also go deeper on the individual components:

The Claude ecosystem is moving fast. Routines are worth getting hands-on with now — before they become table stakes.

Sources: SiliconANGLE — Anthropic's Claude Code gets automated 'routines' and desktop makeover (April 14, 2026)