"A scheduled task runs a prompt on a recurring cadence using Anthropic-managed infrastructure. Tasks keep working even when your computer is off." — Anthropic, 2026.03.27
The latest news from Claude. What if an AI agent could handle your fixed daily tasks at a scheduled time?
Imagine starting your day with a cup of coffee while reviewing a report of your daily tasks, a summary of news related to your interests, and a rundown of unfinished work from the previous day.
That would be a pretty useful assistant — or rather, an agent, right?
In fact, Google Gemini has already been offering such assistant services by leveraging Google's Task feature. This can be seen as Google's strategy to lock users into its ecosystem by encouraging them to use its multi-application Workspace, even before fully diving into AI. They now use the Gemini AI model to handle daily task reporting. In this regard, Google's ecosystem is remarkably extensive, with its various features deeply woven into our everyday lives.
However, how many users truly master Google Workspace and apply it to their daily lives? How many people primarily use Google Docs instead of Word, or Google Sheets instead of Excel? Fewer than you might think.
What if, while enjoying your morning coffee and writing a blog post, an AI could scrape news in your areas of interest and suggest topics — and then you and the agent could read and discuss the content together, turning that conversation itself into a piece of content? Instead of racking your brain thinking, "What should I write about today?", you could browse interesting articles and exchange perspectives with the agent. Wouldn't that be far more effective? And what if the agent could then publish that article in various formats across all your channels?
What if, for a manager who writes daily reports, the agent could research and provide templates tailored to the target audience?
What if, instead of a simple stock market report, the agent could check trends and scrape news overnight — preparing a comprehensive briefing before the market opens, with only a few key decisions left for the analyst to approve? Wouldn't that dramatically increase productivity?
So, How Does Anthropic's New Feature Differ from Google's Task?
Extracting the core from Anthropic's official documentation:
▶ Three Methods of Claude Scheduled Tasks:
"Cloud tasks — Runs on Anthropic cloud. No machine required. Fresh clone each run."
"Desktop tasks — Runs on your machine. Access to local files. Machine must be on."
"/loop — Runs in CLI session. Inherits from session. Session-scoped only."
If Google Task is about "organizing to-do lists and sending reminders," Claude Scheduled Tasks is about "cloning code repositories, performing actual work, and submitting results as Pull Requests."
In other words, Google is a secretary that tells you, "You need to do this," while Claude is a team member who reports, "I've already done this — please review."
Google Workspace's strength is that it rides on an ecosystem already used by billions. Claude's strength is that it plugs directly into the developer's actual workflow — GitHub, CI/CD, and code review.
"Each run creates a new session alongside your other sessions, where you can see what Claude did, review changes, and create a pull request." — Anthropic
Each execution creates a session. You can see what Claude did, review the changes, and create a PR. This isn't a "notification" — it's actual task execution.
Let's Break It Down Simply
Anthropic has given Claude three "ways of working."
First, Cloud Task — The Core of This Announcement
Claude works on Anthropic's servers even when your computer is off. Every morning at 9 AM, it reviews yesterday's code, fixes issues, and submits a PR. You just grab your coffee and check, "What did Claude do?"
The limitation: it can't access your local files. Each run starts from a fresh GitHub clone. This makes it ideal for repetitive tasks like code review, documentation sync, and dependency audits.
Second, Desktop Task
This runs on your machine. It can access local files — manipulate project folders, control browsers, query local databases. But your computer must be on.
Third, /loop — In-Session Repeat
You tell Claude, "Check this every 5 minutes" during a conversation. When the session ends, the loop ends. Lightest weight, shortest lifespan.
The bottom line: If Google is "a secretary who tells you what to do," Claude is "a team member who does the work and reports back." Even while you sleep.
Wrapping Up
Which AI agent do you share your daily life with? As a Claude Max (20x) subscriber, I expect this feature to push productivity even further when running the web and CLI versions of Claude Code at full power.
Anthropic's pace of evolution is truly remarkable. As a developer building MONKOS AI and a photo studio owner for over 20 years, this aligns perfectly with my workflow. AI will handle the routine tasks that need to happen at the exact same time every day.
Recently, MONKOS AI's multi-agentic system has become remarkably sophisticated — agents collaborating and executing work in concert. With more routine tasks emerging from this structure, the timing of this feature couldn't be better.
I'll share a deeper review once I've put it through its paces in real work. Stay tuned.
Questions and feedback are always welcome. Visit us at MONKOS AI and share your thoughts — we'd love to hear from you.