Claude Code Routines and the Access Problem
Claude Code's new Routines feature ran my Google Calendar briefing immediately, no setup. The reason: authentication was already there.
The Step That Usually Blocks
Setting up an automation that reads your calendar involves the same step almost every time: authentication. OAuth flow, credential storage, a token that needs refreshing, a connector that needs configuring. The task itself takes minutes. Getting access takes longer.
I've written before about the delegation problem of the internet: most digital systems assume one account, one person, one operator. Automation quietly breaks on that assumption.
Routines
Claude Code recently added a scheduling layer called Routines, currently in research preview. A routine is an automation you configure once, including a prompt, a repo, and your connectors, and then run on a schedule, from an API call, or in response to an event.
As a quick test, I set it up to print my Google Calendar each weekday at 8:00 AM, the same recurring job I already run in n8n.
It worked immediately. No connector setup, no OAuth dance, no credential to store. Claude already has access to my Google Calendar through the existing integration. The Routine just used it.
That's the actual story. The scheduling feature is convenient. The inherited access is the payload.
Familiar Ground
n8n still does more. The workflow logic is deeper, the integrations more composable, the control more granular. For anything complex, that matters. But for tasks that already live inside the Claude environment, Routines removes the step that most often kills the momentum to automate something in the first place.
There's a category of small, recurring jobs that never get automated because the setup cost exceeds the value: a daily briefing, a weekly summary, a reminder that pulls live data. The barrier is rarely the task itself. It's the plumbing. When the plumbing is already in place, the calculus shifts.
Still Early
There are limits to factor in. On Claude Pro, Routines allows 5 runs per rolling 24 hours, with additional runs drawing on Extra Usage. Max subscribers get 15. For a single daily briefing, 5 is enough. For anything more frequent, the economics need checking first.
My runs haven't visibly registered in the usage counter yet, so how the accounting actually works in practice is still unclear to me. Scheduled tasks also have a way of failing quietly, and reliability over weeks matters more than a clean first run. That's where dedicated tools earn their place.
But the direction is legible. Authentication and access, the layer that blocks most automation for most people, are becoming inherited properties of the tools where the work already happens. The Routine ran. The interesting part was what I didn't have to do first.

