Anthropic just eliminated your cron jobs. Claude Code Routines launched April 14, 2026 as a research preview: scheduled automations running on Anthropic’s cloud infrastructure. Your Mac doesn’t need to stay online. You don’t manage cron jobs or MCP servers anymore. Anthropic handles the infrastructure. It’s GitHub Actions for AI coding, except it’s proprietary and costs $100 per month for team seats.
What Routines Actually Does
Claude Code Routines are scheduled automations that run on Anthropic’s servers, not your local machine. You package up workflows—data cleanup scripts, API integrations, GitHub PR reviews—and set them to run on a schedule or trigger. Anthropic executes them in the cloud.
The daily limits depend on your subscription tier. Pro users ($20/month) get 5 routines. Max subscribers ($100/month) get 15. Team and Enterprise accounts get 25. Hit your limit? Upgrade or delete routines.
What can you automate? The official use cases are scheduled tasks, API workflows, and GitHub-based routines. A practical example: “Read contacts in this CSV, deduplicate by email, normalize phone numbers to E.164 format, write cleaned data.” Claude Code writes the script, runs it, shows results. No manual coding required.
Routines replaces three things: cron jobs for scheduling, self-hosted MCP servers for tool connections, and custom automation infrastructure. Anthropic bundles it all into one cloud service.
The Vendor Lock-In You’re Buying
Routines only works inside Claude Code. Your automation logic isn’t portable to Cursor, GitHub Copilot, or other AI coding tools. You’re locked into Anthropic’s infrastructure.
The pricing escalates fast. Pro gives you 5 routines for $20/month. Teams need Team Premium at $100 per seat per month. Five developers? $500/month. Ten? $1,000/month. Minimum five seats required.
Anthropic dropped Team Premium from $150 to $100 per seat in January 2026. But pricing isn’t fixed. What happens when they increase it?
The migration risk is real. If you build critical workflows on Routines and later switch tools, you’re rewriting everything. Your cron logic, scheduling configs, automation scripts—all tied to Claude Code’s proprietary platform.
The Alternative That Already Exists
GitHub Actions already does scheduled automation with cron expressions, secrets management, log management, and error notifications. It’s free for public repositories. Most development teams already use it for CI/CD.
Anthropic built an official GitHub Action called anthropics/claude-code-action that runs Claude Code inside GitHub Actions runners. You get Claude’s AI capabilities with GitHub’s scheduling infrastructure. The automation logic stays portable.
Before Routines launched, Anthropic’s own documentation recommended: “For durable scheduling that survives restarts, use Cloud or Desktop scheduled tasks, or GitHub Actions.” They told developers to use GitHub Actions. Now they’re selling a proprietary alternative to their own recommendation.
This is the paradox. Anthropic created MCP (Model Context Protocol) as an open standard so AI models can connect to external tools. MCP is open and tool-agnostic. But Routines is closed, proprietary, and Claude Code-exclusive. They’re betting developers will choose native integration over openness.
Who Should Actually Use This
Routines makes sense for specific use cases. If your team is deep in the Claude Code ecosystem and values convenience over portability, it’s reasonable. If you’re a small team and 5-15 routines cover your needs, the pricing works.
But Routines is a bad fit if you use multiple AI coding tools, value infrastructure control and portability, are budget-conscious ($100/seat adds up), or already have GitHub Actions workflows that work.
Claude Code hit $2.5 billion in annualized revenue by early 2026. Forty-six percent of developers named it their most loved tool. Routines is Anthropic doubling down by offering tighter integration at the cost of lock-in.
What This Signals About AI Coding
AI coding tools are shifting from local execution to cloud infrastructure. GitHub Copilot runs suggestions in the cloud. Cursor offers cloud-based pair programming. Now Claude Code provides cloud-hosted automation. Infrastructure-as-service is becoming the default.
If Routines succeeds, every AI coding tool will add similar proprietary infrastructure. If it fails, open standards like MCP and GitHub Actions will remain dominant. The outcome depends on whether developers value convenience enough to accept vendor lock-in.
Routines solves a real problem. Managing cron jobs, MCP servers, and custom automation infrastructure is friction. But GitHub Actions already solves the same problem without lock-in. You can run Claude Code via anthropics/claude-code-action and keep your automation portable. The question isn’t “should I automate?” It’s “should I trade portability for tighter integration?” Your answer depends on how deep you’re willing to go into Anthropic’s ecosystem. Routines is convenient. It’s also a one-way door.
