# Building AI automations with n8n

> n8n is a workflow automation platform you can run in n8n's cloud or self-host, with hundreds of built-in nodes, an HTTP Request node for any API, JavaScript and Python code steps, and an AI Agent node that works with OpenAI, Anthropic and other providers. It is the best fit when workflows are long, high-volume or AI-heavy, or when data must stay on your infrastructure. It is a weaker fit for non-technical teams who want to build their own simple automations.

- Category: Automation & AI Platforms
- Vendor: https://n8n.io/
- How to connect: n8n Cloud; Self-hosted with Docker or Kubernetes; Webhook and schedule triggers; MCP Server Trigger; n8n public API
- Authentication: Credentials stored encrypted per instance, including OAuth 2.0, API keys and header auth; SSO on business and enterprise plans, with role-based access that deepens on higher plans
- Notes: Self-hosted n8n is licensed under the Sustainable Use License, which allows internal business use but is not an OSI open-source license. Queue mode with workers is the standard setup for higher volume.
- Last updated: 2026-10-14

## What you can automate with n8n

- **AI invoice processing.** Read invoices from email, extract fields with a model, validate in code and create bills in QuickBooks or Xero. Guide: https://www.lumeor.studio/guides/automate-invoice-processing-with-ai/
- **Lead qualification agent.** An AI Agent node researches each inbound lead with tools, scores it and writes the result to the CRM. Guide: https://www.lumeor.studio/guides/ai-lead-qualification-agent/
- **Shared inbox triage.** Classify incoming email, draft replies and route each message to the right person or queue. Guide: https://www.lumeor.studio/guides/ai-email-inbox-triage/
- **RAG assistant over company documents.** Index documents into a vector store and answer questions in Slack or a chat widget with sources cited. Guide: https://www.lumeor.studio/guides/build-rag-knowledge-assistant/
- **Data sync between apps.** Keep CRM, billing and support records in step with deduplication, retries and an error queue. Guide: https://www.lumeor.studio/guides/automate-data-sync-between-apps/
- **Alert and log summaries.** Collect alerts from monitoring tools and post a short, prioritized summary instead of a flood of messages. Guide: https://www.lumeor.studio/guides/ai-log-and-alert-summaries/
- **Weekly operations reporting.** Pull numbers from several systems every Monday and post a written summary with the changes that matter. Guide: https://www.lumeor.studio/guides/automate-weekly-ops-reporting/

## What n8n is good for

n8n sits between no-code tools and custom software. You build workflows on a visual canvas, but every step exposes its raw JSON, expressions are JavaScript, and you can drop a JavaScript or Python code step anywhere. That makes it comfortable for technical operators and engineers who want speed without giving up control.

It is strongest in four situations:

- **Long or branching workflows.** Branches, loops, merges, error workflows and sub-workflows are first-class. A 30-step process stays readable.
- **High volume.** On paid plans, n8n bills per workflow execution, however many steps the run has, and the self-hosted Community Edition has no execution fees. Long workflows run many times a day cost far less than on per-step billing.
- **AI-heavy work.** The AI Agent node, chat triggers, vector store nodes and support for many model providers make it a practical place to build agents that touch business systems.
- **Data control.** Self-hosting puts the workflows, credentials and execution data in your own cloud account or data center.

n8n can also act as an MCP server. The MCP Server Trigger exposes selected workflows as tools that Claude, ChatGPT or other MCP clients can call, and the MCP Client Tool lets an n8n agent use outside MCP servers.

## How Lumeor uses it

n8n is our default platform for client automation. A typical build has a trigger, a normalization step, an LLM step with a strict JSON schema, validation in code, the writes to business systems, and an error branch that posts to Slack or Teams with a link to the failed execution. Human approval steps use n8n's send-and-wait operations in Slack, Teams or email, so a person approves anything that moves money or contacts a customer.

Our [n8n automations suite case study](/case-studies/n8n-automations-suite) shows this approach applied to messaging and intelligence workflows on Telegram and WhatsApp. For how we scope, build and maintain these projects, see our [n8n services page](/n8n/).

## Limits to plan around

- **You own the instance.** Self-hosting means upgrades, backups, monitoring and security patches. Budget time for them or use n8n Cloud.
- **Scaling takes setup.** High volume needs queue mode with Redis, separate workers and a Postgres database, not the default single process.
- **Fewer prebuilt connectors.** n8n has hundreds of nodes, not thousands. Niche apps go through the HTTP Request node, which is fine for engineers and harder for business users.
- **Testing and versioning.** Workflows are JSON. Git-based source control and environments exist on higher editions; on others, export workflows and keep them in a repository yourself.
- **License terms.** The Sustainable Use License allows internal use but restricts offering n8n itself as a service to others. Read it if you plan to embed n8n in a product.

## When to pick something else

- **Zapier** for short automations built by non-technical teams across a very wide app catalog. See [n8n vs Zapier](/compare/n8n-vs-zapier/).
- **Make** for business users who need routers and iterators in a polished visual builder. See [n8n vs Make](/compare/n8n-vs-make/).
- **Power Automate** for Microsoft 365 companies that need tenant-governed flows or desktop automation.
- **Code** for agents with complex state, long-running tasks or product-grade testing, often built directly on the [Claude API](/tools/claude-api/) or the OpenAI API.

## Frequently asked questions

### Is n8n free?

The self-hosted Community Edition has no per-execution fees, but you pay for hosting and you own upgrades, backups and uptime. n8n Cloud and the business and enterprise editions are paid and billed by workflow executions. Check n8n's pricing page for current plans.

### Can n8n run real AI agents?

Yes. The AI Agent node runs a tool-calling loop with a chat model, memory and tools, where tools can be other nodes, sub-workflows, HTTP calls or MCP servers. For agents that run for a long time, need complex state or must be unit tested like software, a code framework is often the better home.

### Is n8n secure enough for customer or financial data?

It can be. Self-hosting keeps data inside your own cloud account, credentials are encrypted at rest, and business editions add SSO, role-based access and audit features. Security still depends on how the instance is run, so patching, network access and backups need an owner.

### When should we not use n8n?

When the people building automations are non-technical and the workflows are short, Zapier is easier. When the company runs on Microsoft 365 and IT wants everything governed in one tenant, Power Automate fits better. When the agent is the product itself, write it in code.

### Does Lumeor host n8n for clients?

We set up n8n in the client's own cloud account or on n8n Cloud, so the client owns the instance and the workflows. We can maintain it afterwards, but nothing is locked to us.

---

Source: https://www.lumeor.studio/tools/n8n/
Published by Lumeor Studio, Newtown, PA. hello@lumeor.studio. Get a written scope: https://www.lumeor.studio/start/