n8n

What Is n8n in March 2026?

n8n is an open-source workflow automation platform. But that description sells it short. In March 2026, it’s become an AI-powered orchestration tool — combining a visual workflow editor, 500+ integrations, built-in AI agent nodes, and MCP (Model Context Protocol) support in a package you can self-host for free.

I’ve self-hosted n8n on a $10/month Hetzner VPS for 7 months. Over 40,000 executions later, here’s what I’ve learned: n8n isn’t trying to be the easiest. It’s trying to be the most flexible. And if you’re technical enough to use it, nothing else matches the value.

Why Does Self-Hosting Matter?

Because you own everything. Your workflows, your data, your execution history — all on your infrastructure, with no vendor lock-in, no usage caps, and no surprise bills that scale with complexity.

I moved from Zapier to n8n after a $340 bill for a month where my workflows ran more than expected. On n8n Community Edition, those same workflows cost me $10/month in server hosting. Total.

The trade-off is real, though. You’re responsible for updates, backups, SSL certificates, and monitoring. If your VPS goes down at 2 AM, that’s your problem. Managed platforms handle all of that for you.

How Does the Execution-Based Pricing Work?

One workflow run = one execution. Doesn’t matter if your workflow has 3 steps or 50 steps. It counts as one execution on n8n.

Compare that to per-task platforms, where every step inside a workflow counts as a separate “task.” A 10-step Zap that runs 100 times uses 1,000 tasks. On n8n, that’s 100 executions.

For simple workflows, the difference is small. For complex, multi-step automations, n8n’s model saves serious money.

What Are the Key Features in March 2026?

Visual Editor + Code

n8n’s canvas gives you drag-and-drop workflow building. But unlike drag-and-drop-only tools, you can drop into JavaScript or Python inside any node. This hybrid approach means you’re not stuck when the visual editor can’t handle edge cases.

For debugging, you can re-run single steps, replay data, and inspect outputs at every node. It’s closer to a development environment than a drag-and-drop tool.

AI Agent Nodes + MCP

This is where n8n pulled ahead in early 2026. Native AI agent nodes let you build autonomous workflows that reason, decide, and act. LangChain integration means you can orchestrate Claude, Gemini, GPT, or any LLM within a single workflow.

The March 2026 MCP updates are significant: n8n now acts as an MCP server, letting external AI tools (like Claude Desktop) create, execute, and manage n8n workflows directly. The AI Workflow Builder generates draft workflows from plain English prompts.

I’ve built an AI agent that monitors Slack for support tickets, categorizes them using Claude, creates Jira tasks, and assigns them — fully autonomous. That’s 4 integrations and 12 steps for $0/month on self-hosted.

v2.0 Task Runners

Released in late 2025, Task Runners isolate workflow execution into separate containers. This matters for security — a malicious or buggy workflow can’t affect others. Enterprise-grade isolation without enterprise-grade pricing.

500+ Integrations

n8n supports 500+ native nodes. That’s less than the 7,000+ you get on premium platforms, but n8n’s HTTP Request node connects to any API. If a service has an API, you can connect it. Period.

The difference: Per-task platforms give you pre-built connectors for everything. n8n gives you connectors for the popular ones, and raw API access for everything else. Developers don’t mind. Non-technical users do.

Where Does n8n Fall Short?

Let’s be specific about the gaps.

The learning curve is real. Setting up n8n on Docker took me about 3 hours including SSL and reverse proxy. Building my first production workflow took another 2 hours. On a managed platform, I had something running in 15 minutes. If you aren’t comfortable with servers and APIs, n8n isn’t for you.

Cloud Starter caps at 2,500 executions. Sounds like a lot until you have 10 workflows running on schedules. I burned through 2,500 in the first week when testing. You’ll want Pro ($50/month) or self-hosted.

Community Edition misses enterprise features. No SSO, no Git version control, no environment management. Those require Business ($667/mo) or Enterprise. If you’re on a 5-person team and need SSO, the pricing jump hurts.

Documentation has gaps. Core concepts are covered well. But advanced patterns — complex error handling, nested sub-workflows, API authentication edge cases — often send you to the community forum. The 50K+ member forum is active, though. You’ll usually get answers fast.

n8n vs. the Competition

• vs. Zapier: The managed platform wins on ease of use, integration count (7,000+), and zero technical requirements. n8n wins on flexibility, pricing, self-hosting, and AI agents. Choose Zapier if you’re non-technical. Choose n8n if you develop.
• vs. Make: Make has a cleaner visual interface and operation-based pricing that’s cheaper for simple workflows. n8n’s execution-based billing wins for complex workflows (many steps). Make is more polished. n8n is more powerful. See our Make vs Zapier comparison for context.
• vs. Activepieces: Activepieces is the closest open-source alternative. Simpler interface, fewer integrations, less mature AI capabilities. If n8n feels too complex, try Activepieces.

Who Should Use n8n?

• Developers and DevOps — self-host for free, write custom code inside workflows, version control with Git (Business+)
• Cost-sensitive teams — running 10K+ tasks/month on competing platforms gets expensive fast. n8n’s model saves 50-80%
• AI builders — native automation nodes + March 2026 MCP integration means n8n is becoming the go-to AI orchestration layer
• Regulated industries — self-hosted = full data sovereignty. Enterprise adds SOC 2, RBAC, and audit logging

Who should skip n8n? Non-technical users who want something working in 15 minutes, teams without someone who can manage a server, or anyone who needs 7,000+ pre-built connectors out of the box. For those use cases, Zapier is still the safer choice. See how n8n ranks in our Best AI Automation Tools 2026 list.

Our Verdict

n8n isn’t the easiest automation tool. It’s the most liberating one.

After 7 months and 40,000+ executions on a $10 VPS, I can’t imagine going back to per-task billing. The AI automation capabilities alone — building autonomous workflows that reason and act across multiple services — put n8n in a category competitors haven’t entered yet.

Start with the Community Edition. If Docker and VPS management feel natural, you’ll save thousands per year. If they don’t, start with Cloud Pro ($50/mo) and let n8n handle the infrastructure while you focus on building. Both Cloud plans offer a 14-day free trial.

Pricing verified March 27, 2026. Sources: n8n.io, n8n.io/pricing, github.com/n8n-io/n8n, community.n8n.io.