Node-RED vs n8n
Both are visual, Node.js-based automation tools you can self-host for free, and both let you drop into JavaScript when the boxes run out. They were built for different jobs, though: one to wire hardware, MQTT and home automation on a Raspberry Pi, the other to connect SaaS APIs and AI models with as little code as possible.
n8n vs Node-RED: which should you self-host?
Self-host Node-RED if your flows touch hardware, MQTT, Home Assistant or anything on the edge: it is Apache 2.0, idles at roughly 100 MB and runs on a Raspberry Pi. Self-host n8n if your flows connect SaaS APIs, databases and AI models: 400+ integration nodes, an AI Agent node, credential management, and a $7/mo pod on InstaPods with executions unmetered.
The one-paragraph verdict
Most people comparing these two are really choosing between two jobs. Node-RED is flow-based programming: you wire messages between nodes, write JavaScript in a Function node when you need logic, and it excels at MQTT, serial ports, GPIO, dashboards and event plumbing on small hardware. It is an OpenJS Foundation project under Apache 2.0, currently at version 5.0, and it runs happily in the memory a Pi has to spare. n8n is SaaS-integration automation: 400+ built-in nodes with OAuth credentials handled for you, a Code node that runs JavaScript or Python, an AI Agent node with memory and tools, and per-node execution history for the run that failed at 3am. It ships under the Sustainable Use License, which is fair-code and source-available, not open source, and it wants a real server, around 2 GB of RAM. If the word MQTT appears in your plan, pick Node-RED. If the words Slack, Notion, Postgres and OpenAI do, pick n8n. On InstaPods, n8n is a 1-Click app on the $7/mo Build plan; Node-RED is not in the catalog, though it installs in one command on a plain Node.js pod.
Last updated: 2026-09-14Quick Comparison
| Feature | Node-RED | n8n |
|---|---|---|
| Licence | Apache 2.0 (OSI open source) | Sustainable Use License (fair-code, source-available) |
| Governance | OpenJS Foundation project, started at IBM | n8n GmbH, venture-backed |
| Current Version | 5.0.7 (September 2026), needs Node.js 22.9+ | 2.38.7 (September 2026) |
| GitHub Stars | 23,650 | 204,204 |
| Idle RAM | Roughly 100 MB; the Raspberry Pi guide caps the heap at 256 MB | Around 2 GB for a real instance (n8n publishes no minimum) |
| Built For | IoT, MQTT, edge devices, home automation, event wiring | SaaS APIs, databases, AI agents, business workflows |
| Integrations | MQTT, HTTP, TCP/UDP, serial and GPIO nodes built in; 6,000+ nodes in the flows.nodered.org library | 400+ built-in nodes with OAuth credentials, 900+ community nodes |
| Code Inside a Flow | Function node: JavaScript against the msg object | Code node: JavaScript and Python |
| AI Agents | Community nodes for OpenAI, Ollama and MCP; nothing first-party | AI Agent node with memory, tools and vector stores; MCP server and client |
| Credential Management | Per-node config; secrets in settings.js or environment variables | Encrypted credential store with OAuth flows built in |
| Execution History | Debug sidebar and logs; no per-run history | Per-node execution history with replay |
| Hosted Version | None first-party; third parties such as FlowFuse sell managed Node-RED | n8n Cloud, Starter €20/mo billed annually for 2,500 executions |
| On InstaPods | Not in the catalog; npm install on a Node.js pod | 1-Click app, $7/mo Build plan, executions unmetered |
Where Node-RED Runs Out
Trade-offs to weigh before committing.
You Build the SaaS Integrations Yourself
Node-RED ships the transport layer: HTTP, MQTT, TCP, WebSocket. Talking to Slack, HubSpot or Google Sheets means finding a community node of varying quality or wiring the REST API by hand with an HTTP Request node and a Function node. n8n has an authenticated node for each, with the OAuth dance already done.
No Credential Store
Node-RED has no central, encrypted credential manager. API keys live in individual node configs, settings.js or environment variables, and there is no built-in OAuth handling. Fine on a Pi in your cupboard; awkward once ten workflows share the same three tokens.
Debugging Is Live, Not Historical
The debug sidebar shows messages as they flow. What it does not give you is a stored record of every past run with the input and output of each node, which is the thing you need when a scheduled job failed overnight. n8n keeps per-node execution history and lets you re-run a failed execution.
AI Is Bolted On
There are community nodes for OpenAI, Ollama and MCP, and they work. There is no first-party agent abstraction, no memory or tool sub-nodes, and no vector-store integration. If the workflow is an AI agent with a dozen tools, n8n has spent two years building exactly that.
What n8n Gives You Instead
What you get when you pick this stack.
$7/mo, Executions Unmetered
n8n on InstaPods runs on the Build plan: 2 vCPU, 2 GB RAM, 25 GB SSD, HTTPS, daily backups and a managed PostgreSQL on the same pod. No execution counter, no per-task price. The 2 GB is the whole point of the plan floor: n8n needs it, Node-RED does not.
400+ Integrations With Credentials Handled
Slack, Notion, Postgres, Stripe, GitHub, Google Sheets and hundreds more, each with an authenticated node and the OAuth flow built in. The HTTP Request node covers anything without one, and the Code node takes JavaScript or Python inline.
AI Agents and MCP, First-Party
The AI Agent node has memory, tool and vector-store sub-nodes. n8n can be an MCP server (instance-level or per-workflow) and an MCP client, so Claude can call your workflows and your workflows can call other MCP servers. See [connecting n8n to Claude over MCP](/blog/connect-n8n-to-claude-mcp/).
Execution History You Can Replay
Every run is stored with the input and output of every node. When the 6am sync fails, you open that execution, see which node broke and on what data, fix it and re-run from there. Node-RED shows you the live stream; n8n shows you the past.
Real Cost Comparison
A developer automating a small business: a daily Postgres report to Slack, a webhook from Stripe into a CRM, and an AI step that summarises support emails. Self-hosted either way.
- Node-RED: free (Apache 2.0)
- Any small VPS or a Node.js pod: $3-10/mo
- Slack, Stripe, CRM: community nodes or hand-wired HTTP calls
- OAuth tokens: managed by you in settings.js
- Failed-run history: none; add your own logging
- InstaPods Build plan: $7/mo (2 vCPU, 2 GB, 25 GB)
- n8n community edition: free, executions unmetered
- Slack, Stripe, Postgres, OpenAI nodes: built in, OAuth handled
- Managed PostgreSQL, HTTPS, daily backups: included
- Per-node execution history and replay: included
When Node-RED is the right call
If the job is hardware, Node-RED wins outright. MQTT in and out, serial ports, GPIO on a Pi, Modbus, Zigbee bridges, a dashboard for the sensor readings: these are first-class in Node-RED and an afterthought in n8n. It runs as a Home Assistant add-on, it has been the default glue in home automation for a decade, and its memory footprint means it lives on the same $35 board as the sensors. n8n at 2 GB does not belong on that board.
The licence matters too. Apache 2.0 lets you embed Node-RED in a product you sell and host it for customers. n8n's Sustainable Use License permits internal business use and non-commercial distribution, but not offering n8n itself to third parties as a paid service. If you are building a product on top of the automation engine rather than automating your own business, that difference decides it. For a longer list of permissively licensed options, see the nine self-hosted n8n alternatives.
Frequently Asked Questions
n8n vs Node-RED: which should you self-host?
Self-host Node-RED for hardware, MQTT, Home Assistant and edge work: it is Apache 2.0, idles at roughly 100 MB and runs on a Raspberry Pi. Self-host n8n for SaaS and AI automation: 400+ integration nodes with OAuth handled, a Code node in JavaScript or Python, an AI Agent node and per-node execution history. On InstaPods, n8n is a 1-Click app at $7/mo; Node-RED is not in the catalog.
Is Node-RED an n8n alternative?
For SaaS-style automation, only partly. Node-RED can call any REST API through its HTTP Request and Function nodes, so it can technically do what n8n does, but you build each integration yourself and manage credentials by hand. It is a genuine alternative if your workflows are event plumbing, MQTT or hardware, or if you need a permissive licence and a tiny footprint. It is a poor alternative if you want 400 pre-built SaaS nodes and an AI agent framework.
Is n8n open source?
No. n8n is source-available under the Sustainable Use License, which n8n describes as fair-code. You can read, modify and self-host it, and use it for internal business or personal purposes for free, but you may only distribute it free of charge for non-commercial purposes and cannot offer it to third parties as a paid hosted service. Node-RED is Apache 2.0, an OSI-approved open-source licence with no such restriction.
Which uses less RAM, Node-RED or n8n?
Node-RED, by a wide margin. It idles at roughly 100 MB, and the official Raspberry Pi guide starts it with a 256 MB heap cap. n8n publishes no minimum, but its own Hetzner guide points at a 2 GB box and the AI Assistant needs 4 GB. That is why n8n on InstaPods starts on the 2 GB Build plan at $7/mo rather than the $3/mo Launch plan.
Can Node-RED and n8n work together?
Yes, and it is a common setup. Node-RED handles the device layer, publishing sensor events over MQTT or calling a webhook, and n8n picks those up with its Webhook or MQTT Trigger node to run the business side: write to Postgres, alert Slack, hand the data to an AI model. Each tool does the part it is built for.
Can I run Node-RED on InstaPods?
Not as a 1-Click app. Node-RED is not in the InstaPods catalog. It is a plain npm package, so it installs on a Node.js pod with npm install -g node-red, but you would set up the admin login and the reverse proxy yourself. n8n is the 1-Click app, on the $7/mo Build plan with HTTPS, backups and a managed PostgreSQL included.
Deploy n8n in 60 Seconds
400+ integrations, AI Agent node and execution history, for a flat $7/mo with executions unmetered. Keep Node-RED on the Pi for the sensors.
Deploy n8n