n8n

n8n is the strongest open-source alternative to Make and Zapier, and over 2025–26 it has become something more specific: the de facto standard for self-hosted AI-agent workflows. Its node-based canvas covers the familiar automation territory — triggers, branching, data transformation, roughly 1,200 connectors — and adds first-class AI-agent nodes (LangChain-based, model-agnostic) that treat LLM steps, tool use and memory as native building blocks rather than bolted-on API calls.

Image: n8n.

The licensing and pricing model is the story. The Community Edition is free to self-host with unlimited executions — the whole product, not a teaser — under a fair-code licence (source-available with commercial restrictions; close to, but formally not, OSI open source, and worth stating precisely). n8n Cloud bills per workflow execution: one run of an entire workflow counts once, however many steps it contains, which is materially cheaper than Make's per-module credits or Zapier's per-task metering for complex flows. Cloud starts at €24/month for 2,500 executions with AI credits bundled (verified Aug 2026; annual billing saves roughly 17%).

The honest trade-offs mirror the ERPNext story from our ERP coverage: you exchange licence cost for responsibility. Self-hosting is a real ops job (updates, backups, queue scaling), the learning curve is at least Make's and steeper than Zapier's by a distance, and the connector catalogue — around 1,200 against Zapier's ~7,000 — means long-tail apps sometimes need the HTTP node and your own API reading. Teams without a technically confident owner should buy the managed cloud or stay with the mainstream platforms.

Full hands-on production deployment testing is in progress; this profile reflects n8n's documentation, its published licence terms and our automation-consulting experience with the mainstream platforms it competes against.