What Happened
Anthropic shipped Claude Routines — saved Claude Code configurations that run on a schedule, on an API call, or when a GitHub event fires. Anthropic infrastructure runs the job; your laptop does not have to be online.
If you already use n8n or Make.com for automation, the obvious question is which one you use now. The short answer: three different problems.
The Headline Difference: Natural Language vs Knobs
With Claude Routines, you describe what you want in plain English:
"Every morning at 7am, read the main branch of this repo, check if any dependencies have known CVEs, and if so, open a PR that bumps them safely and runs the tests."
That sentence is the configuration. You save it, Anthropic runs it.
With n8n or Make, the same job is a wiring exercise:
- Auth: OAuth into GitHub, configure scopes, store the token
- Connectors: add a GitHub node for "list dependencies", another for "create PR", another to trigger workflows
- Field mapping: connect step 1's output to step 2's input, JSON-stringify here, base64 there
- Branching: add an If node for "any CVEs?" — true and false paths
- Error handling: error branches, retries, the Slack node that fires on failure
- Scheduling: set the cron expression
Nothing is hard. Collectively, it's an hour of setup before the flow does anything useful.
Where Claude Routines Wins
- Automated code review. Trigger on every PR. Claude reads the diff, the surrounding code, and the tests, then comments with specific issues.
- Release verification. Trigger on deploy. Claude hits the production URL, runs a smoke test, opens an issue if anything looks off.
- Alert triage. Sentry webhook → Routine reads the stack trace, checks git history, posts a root-cause summary.
- Dependency maintenance. Weekly schedule. Claude checks the lockfile, runs tests against bumped versions, opens a safe-upgrade PR.
In every case the signal is: something happened to a repo, do a thing with the repo. Routines are the shortest path.
Where n8n Wins
- Self-hosted, unlimited runs. Automation living entirely inside your infrastructure with no per-operation metering.
- Complex branching and custom code. First-class code nodes (JavaScript or Python) inside a visual flow.
- Engineer ownership. The mental model maps to how engineers already think about pipelines.
- Workflows the business depends on long-term. You own the data, uptime, and upgrade path.
Where Make.com Wins
- Fastest setup. The friendliest onboarding of the three. A sales ops person or marketer can own a flow end-to-end.
- Breadth of connectors. 1,500+ apps, catalog skews toward marketing, ecommerce, and CRM tools.
- Marketing and ecommerce workflows. Lead capture → CRM enrich → email sequence → Slack alert. Make's sweet spot.
- No infrastructure to maintain. Fully hosted. You pay per operation and Make handles everything else.
Use All Three Together
The real answer for most teams at scale is all three, with different jobs.
- n8n or Make is the durable surface that listens to the world — webhooks, schedules, SaaS triggers — and glues systems together. Pick n8n if you want self-host plus power. Pick Make if you want fastest setup plus broadest connectors.
- Claude Routines lives inside the engineering-minded steps (PR reviewer, release checker, dependency bumper) and inside the "needs reasoning" nodes of your visual flows.
A concrete hybrid: Make receives a Shopify order webhook → calls a Claude Routine via API to analyze the order against fraud heuristics in your codebase → writes the verdict back to Make, which routes the order to fulfillment or review.
The Decision Rule
When you need to automate something, ask in order:
- Is this primarily reading or writing a codebase? Claude Routines.
- Do engineers own this, and do you need self-host or complex branching? n8n.
- Does a non-engineer own this, and is it primarily SaaS glue? Make.com.
- Does it need reasoning somewhere in the middle? Claude Routine as a step inside the n8n or Make flow.
What This Means for You
The overlap between these tools is smaller than it first looks — they're complementary, not competing. The orchestrator's job is knowing which interface fits which work, and wiring them together where reasoning matters.
Claude Code Superpowers and Application Monitoring & Self-Healing both walk through this pattern end-to-end.