What Are Claude Routines?
Claude Routines is a saved automation inside Claude Code: a prompt, a set of connected files or repositories, and linked external tools, packaged together and triggered automatically. You define it once — the trigger, the inputs, the output destination — and it runs on Anthropic's cloud infrastructure without you. Your laptop does not need to be on.
On April 14, 2026, Anthropic shipped the Routines feature for Claude Code — and it quietly changes what non-technical AI practitioners can do with the platform. If you have ever built a prompt that works perfectly, but still had to open it, paste your data, and run it manually every morning or every Monday, Routines removes that friction entirely.
As of the April 2026 release, Routines can execute in three ways: on a scheduled time interval (daily, weekly, or custom), via an API webhook call from another system, or in response to a GitHub event such as a new commit or pull request. Once triggered, the Routine runs inside Anthropic's cloud environment with access to whatever repositories, files, and connectors you have attached to it.
This is meaningfully different from simply saving a prompt template. A saved prompt still requires you to open Claude, paste your input, and evaluate the output. A Routine removes all three steps. The prompt runs, processes the connected data, and delivers output to wherever you have directed it — a Slack channel, an email thread, a file, or a downstream tool.
Who Is Claude Routines Actually Built For?
Claude Routines was originally a developer feature but has been explicitly repositioned for non-technical users. Anthropic's April 14 announcement, covered by 9to5Mac and SiliconAngle, described the goal as "turning tools originally built for engineers into something non-technical people can use to take repetitive work off their plate." No coding is required to configure or run a Routine.
For AI practitioners — marketers, content creators, operations managers, project managers, educators, and freelancers — Routines is particularly well-suited to any task that follows a consistent pattern: same input type, same transformation, different data each cycle. A weekly competitive summary, a daily content brief review, a client-facing report compiled from a shared spreadsheet — these are precisely the use cases Routines was designed for.
The subscription requirement matters: Routines is available to Claude Code Pro, Max, Team, and Enterprise subscribers. Pro users get five Routine runs per day; Max users get 15; Team and Enterprise users get 25. The Pro tier handles most personal workflow automations without hitting the daily ceiling. If you are on the free Claude tier, this feature is not yet available.
What Tasks Can You Automate With Claude Routines?
Routines works best for tasks with a defined input, a consistent transformation, and a predictable output destination. The strongest use cases share a common structure: structured data in, structured summary or draft out, delivered to a fixed location on a reliable schedule.
Here are five concrete workflow categories where Routines delivers immediate value for non-developers:
--- Weekly competitive digest. Connect Claude to web search, write a prompt that scans for competitor activity from the last 7 days, formats it into a structured digest, and posts it to a Slack channel. Schedule it every Monday at 8:00 AM. This replaces 30–45 minutes of manual research with zero manual effort after setup.
--- Document monitoring. When a new PDF drops into a shared folder via a webhook trigger, a Routine can summarize its key points, tag it by topic, and post a summary to your team channel — eliminating the need to manually read and route every incoming document.
--- Content brief execution. Feed Claude a campaign brief document once per week and have it generate draft outlines, headline options, and keyword suggestions for your content team. Schedule it to arrive before your Friday planning session.
--- Data report synthesis. Connect Claude to a Google Sheet or CSV export. Write a prompt that extracts key metrics and formats them into an executive summary. Schedule it to run before your Monday standup meeting.
--- Email draft generation. Connect Claude to a filtered inbox label. Set a Routine to scan flagged emails and generate reply drafts in your established tone, stored in a draft folder for your review.
The key constraint: Routines execute a task defined at setup time. If your workflow requires dynamic mid-task decision-making — or judgment calls that depend on real-time context outside the connected data — manual oversight is still needed for those steps.
How Do You Set Up Your First Claude Routine?
Setting up a Routine in the April 2026 redesigned Claude Code interface takes three core steps: write a task prompt, connect your data sources, and define the trigger. The entire configuration lives in the Routines section of the left sidebar, and no code is written at any stage.
Here is the exact setup flow as of the current interface:
--- Step 1. Open Claude Code and click "Routines" in the left sidebar. Click "New Routine."
--- Step 2. Write your task prompt in plain English. This is the core instruction Claude executes each time the Routine runs. Be specific: include output format, target length, tone, and how Claude should handle ambiguous or incomplete inputs.
--- Step 3. Attach repositories or files this Routine needs to read. If your Routine processes a weekly data export or references a style guide, connect it here.
--- Step 4. Add connectors for external outputs — Slack, email, Google Sheets. This is where the Routine delivers its result.
--- Step 5. Set the trigger. Choose "Schedule" for a recurring time interval, "API" for a webhook URL you can call from Zapier, Make, or n8n, or "GitHub Event" for code repository triggers.
--- Step 6. Run a manual test. Verify the output matches your expectations before enabling the live trigger.
One practical note from coverage of the launch: Anthropic requires the model to be web-enabled for Routines to function. If your organisation's Claude Code account has web access restricted by an admin policy, confirm this setting before building your first Routine.
Where Does Claude Routines Fall Short Right Now?
No feature ships at full capability on launch day. Knowing the current limitations before you build helps you design Routines that will work reliably — and avoid investing setup time in use cases the feature does not yet support well.
The daily run limit is the most common ceiling. Pro users max out at five Routine executions per day. If you need multiple Routines running throughout the day — a morning content brief, a midday data check, an end-of-day summary — the Pro tier caps out quickly. Max tier (15 runs/day) handles most individual power-user workflows, and Team/Enterprise (25 runs/day) suits team-wide automation needs.
Complex real-time web data is less reliable than static files. Routines with web-enabled prompts work well for structured lookups — "find competitor announcements from the last 7 days" — but are less reliable for real-time financial feeds, live API data, or sources that require authentication. Static file inputs (Google Drive documents, GitHub repos, CSV exports) produce the most consistent results.
Long document batches have a token ceiling. Routines are subject to the same context window limits as a standard Claude session. Very large document batches may require chunking the input or running multiple Routines in sequence. For most documents under 500 pages, Claude Sonnet 4.6's 1M-token beta window accommodates the full content in a single pass.
Multi-agent coordination is not yet supported within a single Routine. If your workflow requires Claude to spawn sub-tasks or coordinate parallel outputs, Claude Code's multi-session parallel architecture is more appropriate than a single Routine.
Your Starter Template: A Copy-Paste Routine Prompt
The fastest way to experience the value of Routines is a weekly competitive monitoring digest — a use case that requires no custom file connections, just a prompt and a Slack or email connector. This template is configured and running in under 10 minutes.
Routine Prompt (Weekly Competitive Digest) — paste this as your task description:
"Search the web for news, blog posts, product announcements, and pricing changes from [Competitor A] and [Competitor B] published in the last 7 days. For each company, summarise the top 3 to 5 updates in 2 to 3 sentences each. Flag any pricing changes, new feature launches, and partnership announcements at the top of the output. Format the result as a structured digest with four sections: New Features, Pricing Changes, Partnerships, and Other News. Use a neutral, factual tone. Close with one sentence identifying the most strategically significant update from the week."
Trigger: Schedule — every Monday at 8:00 AM HKT
Connector: Your team Slack channel or email inbox
Files attached: None required
Replace [Competitor A] and [Competitor B] with your actual competitor names. Run it once manually to verify the output quality matches your standards, then enable the weekly schedule. What previously took 30 to 45 minutes of manual research every Monday morning now takes zero ongoing time after the initial 10-minute setup.
What Claude Routines Means for Your Workflow
Routines represents a shift in the relationship between AI tools and your daily work. Until now, most AI productivity gains for practitioners have come from getting better at prompting — which remains worth developing. But Routines begins to answer a different question: once you know how to get reliable output from Claude, how do you make that output available without being present every time?
The practitioners who will extract the most value here are those who have already identified their highest-value, most repetitive AI use cases. They have a prompt that works. They run it on a schedule. They just need the infrastructure to stop requiring them to show up for every execution.
懂AI，更懂你 — UD 相伴，AI 不冷. That shift — from AI as a tool you operate to AI as infrastructure that operates for you — is what Routines is built for. For AI practitioners in Hong Kong who are looking to move from casual daily use to genuine workflow integration, this is the feature worth setting up this week.
Ready to Build Your First AI Automation?
Setting up your first Claude Routine is straightforward — but designing a full AI workflow that integrates reliably with your existing tools takes a different level of planning. The UD team 手把手帶你完成每一步 — from mapping your repetitive tasks to configuring connectors, triggers, and outputs that work together as a system.