Claude Code Routines: The Feature That Replaced Half My Automation Stack
Claude Code just launched Routines - scheduled, API-triggered, and webhook-based automations that run in the cloud. Here's what they are, how I'm using them, and why this changes everything for developers who build with AI.

AI & Web Consultant · April 15, 2026

I've been using Claude Code as my daily development partner for months. Custom slash commands, CLAUDE.md files, subagents - the whole setup. It's genuinely changed how I build.
But there was always one gap: everything was local. If my laptop was closed, Claude wasn't working.
Routines just fixed that.
As of April 14, 2026, Claude Code can now run automations in the cloud - on a schedule, triggered by an API call, or responding to GitHub events. No local machine needed. No cron jobs on a server. No duct-taped workflows in Make or Zapier.
This is a big deal. Let me break down what it is, how it works, and how I'm already using it.
What Are Routines?
Routines are Claude Code automations that you configure once and let run repeatedly. Think of them as "set it and forget it" Claude tasks.
There are three types:
1. Scheduled Routines
Run on a cadence - hourly, nightly, weekly. You write the prompt, pick the schedule, point it at a repo, and Claude handles the rest.
Examples: - Every night at 2am: pull the top bug from your issue tracker, attempt a fix, open a draft PR - Every Monday morning: scan last week's merged PRs for documentation drift, post a summary to Slack - Every hour: check your monitoring dashboard for anomalies
2. API Routines
Triggered by an HTTP POST request. You get a custom endpoint and auth token. Any system that can make an HTTP call can trigger Claude.
Examples: - After a deployment: Claude runs smoke tests and posts results - When an alert fires: Claude reads the alert payload, finds the relevant service, and posts a triage summary - When a customer submits feedback: Claude categorizes it and routes it to the right team
3. Webhook Routines
Respond to GitHub events automatically. A PR gets opened, a branch gets pushed, an issue gets created - Claude jumps in.
Examples: - Every PR touching your auth module: Claude flags it and posts a security-focused review - New PR in your Python SDK: Claude automatically ports the changes to your Go SDK and opens a matching PR - Issue labeled "bug": Claude investigates the codebase and posts initial findings
Why This Matters
Before Routines, Claude Code was powerful but reactive. You had to be there, in the terminal, asking it to do things.
Now it's proactive. It works while you sleep. While you're in meetings. While you're on a beach.
For me personally, this replaces a chunk of what I used to build with n8n, Zapier, or custom cron jobs:
- PR review automation - I used to have a GitHub Action that ran basic linting checks. Now Claude does an actual intelligent code review on every PR. Not pattern matching - real reasoning about the code.
- Nightly maintenance - dependency updates, dead code scanning, test coverage checks. Used to be manual or required custom scripts. Now it's a one-line prompt on a schedule.
- Client project monitoring - for retainer clients, I can set up routines that watch their repos and alert me if something looks off. Proactive maintenance without me checking manually.
How to Set Up a Routine
From the Web
Go to claude.ai/code, navigate to Routines, and create one. You'll configure:
- 1Prompt - what should Claude do? Be specific. This is your CLAUDE.md equivalent for the routine
- 2Repository - which repo should it work on?
- 3Connectors - external tools (Slack, Linear, GitHub, etc.)
- 4Trigger - schedule, API endpoint, or webhook event
- 5Filters - for webhooks, which events should trigger it (e.g., only PRs to main branch)
From the CLI
Use the /schedule command in Claude Code to create routines from your terminal.
Usage Limits
| Plan | Routines per day | |------|-----------------| | Pro | 5 | | Max | 15 | | Team | 25 | | Enterprise | 25 |
Each "routine run" counts as one. A nightly routine that runs once = 1/day. A webhook routine that fires 10 times = 10/day.
My Setup Right Now
Here's what I have running after the first day:
Routine 1: Nightly PR Cleanup (Scheduled - 2am) Checks all open PRs in my active projects. If a PR has been open for more than 3 days with no activity, Claude posts a summary of what's blocking it and suggests next steps.
Routine 2: Security Scan on Auth Changes (Webhook) Any PR that touches authentication, authorization, or session handling gets an automatic security-focused review from Claude. It checks for common vulnerabilities, missing validation, and posts findings as a PR comment.
Routine 3: Client Alert Triage (API) When a monitoring alert fires for a client project, Claude reads the alert, correlates it with recent deployments, and posts a triage summary to our shared Slack channel. Saves me from being the first-responder at 3am.
What This Means for Solo Developers and Small Teams
This is where it gets interesting. Routines essentially give a solo developer or a 2-3 person team the automation infrastructure that used to require DevOps engineers and custom tooling.
You don't need: - A CI/CD specialist to set up complex pipelines - A separate monitoring stack with custom alerting logic - Manual code review processes that nobody follows - Weekly "let's clean up the backlog" meetings
Claude handles the boring, recurring, maintenance work. You focus on building.
The shift: Claude Code went from "an AI that helps you code" to "an AI teammate that works autonomously on your codebase." That's a fundamentally different thing.
Getting Started
If you're already using Claude Code:
- 1Go to claude.ai/code or use
/schedulein the CLI - 2Start with one simple scheduled routine - something low-risk like a nightly summary of open issues
- 3Watch what it does. Read the outputs. Tune the prompt.
- 4Add webhook routines for your most tedious recurring tasks (PR reviews, dependency checks)
- 5Graduate to API routines when you're ready to connect external systems
If you're not using Claude Code yet - this might be the feature that makes it worth trying. The combination of local development assistance + cloud-based automation is something no other AI coding tool offers right now.
Available for Pro, Max, Team, and Enterprise plans.
Want help implementing this?
I help B2B companies implement AI solutions that actually move metrics — not science projects. If this guide resonated, let's talk about what it looks like for your business.
Get in touch