n8n + OpenClaw: The Right Way to Do No-Code Workflow Automation

Let's Be Clear: What n8n Is and Isn't

n8n gets compared to Zapier a lot, but that comparison misleads. Zapier is hosted, priced per action, built for non-technical users. n8n is self-hosted, open-source, aimed at people who are willing to configure things themselves.

Better analogy: n8n is a workflow automation engine for technical users. It has a visual interface, but also full code nodes where you can write JavaScript directly. That makes it particularly useful in the "don't need a complete program but do need flexible logic" sweet spot.

OpenClaw's n8n-automation skill connects n8n with AI agents, so you can describe a workflow in plain language, have the agent generate n8n's JSON config, and even trigger workflow execution directly via n8n's REST API.

Installation and Setup

Prerequisite: you need a running n8n instance. Docker is the easy path:

docker run -it --rm   -p 5678:5678   -v ~/.n8n:/home/node/.n8n   n8nio/n8n

Install the skill via ClawHub:

clawhub install n8n-automation

Then configure your n8n instance URL and API key (generated in n8n's Settings → API) in your TOOLS.md:

### n8n
- URL: http://localhost:5678
- API Key: your-n8n-api-key
- Webhooks base: https://your-domain.com/webhook/

Three Real Use Cases

Use Case 1: RSS Aggregation + AI Summaries + Telegram Push

We monitor a dozen AI news RSS feeds. Reading them manually takes forever. Using the n8n-automation skill, we described the requirement:

Create an n8n workflow: every morning at 8am, read updates 
from these RSS feeds: [URLs]. Filter articles containing 
"LLM" or "agent", generate 100-word summaries using Claude, 
send to Telegram channel @sfdlab_news

The agent generates complete n8n workflow JSON: Schedule Trigger → RSS Read → Filter → HTTP Request (Claude API) → Telegram node. Import to n8n, configure API keys, activate.

Real gotcha: RSS Read's "only return new items" feature depends on local storage. Docker restarts wipe this state, causing the workflow to re-send historical articles on every restart. Fix: add a Postgres node to track processed article URLs.

Use Case 2: GitHub PR Notification + Automated Code Review

When someone opens a PR, n8n catches the GitHub webhook, automatically asks Claude for a preliminary review, and posts results to Discord.

Key setup: n8n's Webhook node gives you a URL (format: /webhook/{uuid}). Add this to GitHub's Settings → Webhooks, select the pull_request event.

Workflow: Webhook → Extract PR info → HTTP Request (Claude) → Discord Webhook.

Claude prompt:

You are a code reviewer. Here's the PR diff: {{$json.diff}}
Briefly assess (under 200 words):
1. Potential bug risks
2. Code style issues  
3. Test coverage adequacy

Note: Claude API calls need the HTTP Request node, not n8n's built-in AI node (which only supports certain providers). Add x-api-key in headers, build the Anthropic-format body manually.

Use Case 3: Database Change Monitoring

Critical tables (payments, user_roles) need audit trails and notifications on any change.

n8n's Postgres Trigger node can listen to PostgreSQL's LISTEN/NOTIFY events. First, set up the trigger in the database:

CREATE OR REPLACE FUNCTION notify_payment_change()
RETURNS TRIGGER AS $$
BEGIN
  PERFORM pg_notify('payment_changes', 
    row_to_json(NEW)::text);
  RETURN NEW;
END;
$$ LANGUAGE plpgsql;

n8n's Postgres Trigger subscribes to payment_changes. On change: write to audit log table, send Telegram notification.

What the n8n-automation Skill Actually Does Well

1. Natural language → workflow JSON: Describe the flow, agent outputs importable n8n JSON
2. Workflow debugging: Paste error messages, agent pinpoints n8n node config issues
3. API triggering: Agent calls n8n's REST API directly, triggers workflow execution mid-conversation
4. Node selection: Tell the agent your requirements, it recommends the right node combinations

Limitation: generated JSON sometimes has outdated node versions (n8n updates frequently). Test in a staging workflow before activating in production.

Pairing with Other Skills

• smart-web-scraper: n8n schedules scraping tasks, agent processes extracted data
• self-improving-agent: workflow execution errors auto-reported, agent learns error patterns
• stock-watcher: market changes trigger n8n workflows with custom alert formatting

SFD Lab Note

We run 6 n8n workflows at SFD Lab. The most important one monitors our content publishing pipeline: after each article publishes, n8n automatically checks URL accessibility and SEO metadata completeness, notifying Fox to re-publish if anything's wrong.

n8n has a steeper learning curve than Zapier, but the self-hosted + code nodes combination gives it a much higher ceiling. If you're doing multi-agent collaboration in OpenClaw, n8n is excellent "glue layer"—chaining agent outputs into truly automated pipelines.