Automation Without an Engineering Team
You don't need a developer to automate the boring parts of running a business. n8n is an open-source workflow tool that connects your apps, and Claude is the reasoning layer that reads, writes, and decides. Together they let you build agents that handle real work on a schedule.
What You Can Automate
- Draft product descriptions from a spreadsheet of specs
- Summarize and tag incoming support emails
- Turn new Shopify orders into Slack alerts with a risk score
- Generate weekly performance reports from your ad and sales data
- Reply to routine WhatsApp or email questions with a human review step
Step 1: Get n8n Running
Use n8n Cloud for the fastest start, or self-host with Docker if you want full control and lower cost. Both give you a visual canvas where each node is an app or an action.
Step 2: Add Your Claude API Key
Create an API key from the Anthropic console, then add an HTTP Request node (or the Anthropic node) in n8n. Point it at the Messages endpoint and choose a current model. Store the key in n8n credentials, never in the workflow itself.
Step 3: Design the Trigger
Every automation starts with a trigger:
- Schedule — run every morning at 9am
- Webhook — fire when a Shopify order is created
- App event — a new row in Google Sheets, a new email, a new form entry
Step 4: Send Structured Prompts to Claude
The secret to reliable automations is a tight prompt. Tell Claude exactly what role it's playing, give it the input data from the previous node, and ask for output in a fixed format (JSON works well). For example: "You are a support triage assistant. Classify this email into billing, shipping, or product, and return JSON with a category and a one-line summary."
Step 5: Act on the Output
Feed Claude's response into the next node — post to Slack, update a Shopify tag, write back to a sheet, or send a draft reply. Add a human-in-the-loop step for anything customer-facing until you trust the quality.
A Real Example: Review-to-Content Pipeline
1. New product review lands (trigger) 2. Claude drafts a social caption highlighting the review 3. Draft posts to a Slack channel for approval 4. Approved captions queue in your scheduler
That's a content engine that runs itself.
Guardrails That Matter
- Keep a human review step on anything public
- Log every run so you can debug bad outputs
- Start with one workflow, prove the time saved, then expand
