Anthropic lets Claude Code work while your laptop is closed

The biggest friction with local AI coding tools has always been the tether to your machine. If you want an agent to comb through backlogs or run a heavy test sweep, you had to keep your laptop awake and your terminal open.

Anthropic is cutting that cord. The company introduced Claude Code Routines in research preview, moving repeatable developer workflows entirely to the cloud.

Automations that run without your laptop

A routine bundles three elements into an autonomous job: a prompt, target code repositories, and connectors like MCP servers. Because they execute on Claude Code web infrastructure, nothing depends on your local environment being active.

There are three ways to kick them off. The first is on a schedule. Developers can set tasks to run hourly, nightly, or weekly. For example, Claude can wake up at 2 AM, pull the top bug from Linear, attempt a code fix, and leave a draft PR waiting for morning review. CLI users running /schedule will find their jobs migrated to this cloud setup.

The second path is via direct API calls. Every routine gets a dedicated endpoint and auth token. Teams can wire it straight into Datadog alerts or CI/CD deploy hooks. When an alert fires, Claude can pull the trace, cross-reference recent commits, and draft a triage summary before on-call engineers even open their laptops.

The third trigger handles repository events, starting with GitHub webhooks. Claude spins up an isolated session whenever a PR matches your rules. Crucially, it stays attached to that PR, processing incoming review comments and CI failures as they happen.

Early teams are already testing practical chores: scanning merged PRs weekly to catch outdated documentation, automatically porting SDK updates from Python to Go, and running custom security checklists before human review.

Access is rolling out to Pro, Max, Team, and Enterprise users on the web version of Claude Code. Runs count against standard subscription limits and carry daily caps: 5 per day for Pro, 15 for Max, and 25 for Team and Enterprise, with paid overages available.

Why it matters

This looks like the tipping point where coding agents shift from interactive pair programmers to true background colleagues.

Until now, running asynchronous agents meant stitching together your own cron jobs, hosting scripts, and managing headless environments. Anthropic essentially packaged all that plumbing into a turnkey service. For solo developers and lean teams, it delivers enterprise-grade automation without the infrastructure tax.

Is it worth trying today? Absolutely, especially for chores where the blast radius is low, like flagging stale documentation, drafting initial bug triages, or checking routine PR formatting.

The real friction will be trust. An AI porting code between SDKs or running post-deploy smoke tests sounds great until hallucinated edge cases slip into production. Teams will need airtight test suites before treating these routines as anything more than draft-generators.

Still, by putting triggers, repos, and cloud compute into a single interface, Anthropic is putting pressure on GitHub Copilot to turn repo automation into an effortless default rather than a custom configuration puzzle.