Self-hosted n8n VPS Hosting

n8n is a fair-code workflow automation platform that connects apps, APIs, and databases through a visual drag-and-drop editor or your own code. It comes with more than 400 built-in nodes for services like Slack, Google Sheets, Postgres, OpenAI, and any HTTP endpoint, plus dedicated nodes for large language models and AI agents. Teams use it to automate work across business operations, data engineering, customer support, and AI development, and the same workflow can mix prebuilt nodes with JavaScript or Python wherever extra logic is needed.

n8n is fair-code rather than strictly open-source in the OSI sense. The Sustainable Use License lets you run, modify, and self-host the software for free for internal business use, but it prohibits embedding n8n into a product you resell to outside users. Source code is public on GitHub, and a community marketplace shares custom nodes and workflow templates. For most teams that want to self-host n8n on their own server, the license is permissive enough that the practical effect lines up with open-source.

n8n Cloud is the managed service from the n8n team, where you pay a monthly fee for a hosted instance and the price is metered by execution count. Plans start around 24 dollars a month for 2,500 executions and 60 dollars a month for 10,000. Self-hosted n8n is the same software running on a server you control, with no per-execution fees and no caps on workflows or users. Self-hosting trades a small ops load for unlimited volume and full data control.

n8n uses about 300 to 500 MB of RAM at idle and climbs to 1 or 2 GB during complex executions. PostgreSQL adds another 512 MB to 1 GB on top, depending on the connection pool. That puts comfortable production for a single instance at 4 to 8 GB of RAM, with 16 GB or more for queue mode and heavy webhook traffic. Memory is the main bottleneck, not CPU, so over-providing RAM is the safer side to err on.

n8n holds whole workflow datasets in memory as live JavaScript objects, so the server process has to stay running, not start and stop on demand. Webhook endpoints can only fire when the instance is up, and cron-style schedules trigger from the same process. If the server reboots or the terminal that started n8n closes, any in-flight work stops and any webhook hit during the gap is missed. A VPS with a process manager keeps everything online around the clock.

The self-hosted Community Edition is free software under the Sustainable Use License, with no limits on workflows, steps, executions, or users. The only ongoing cost is the server you run it on, which is why many teams move from n8n Cloud once their execution counts climb. Running n8n on a GreenGeeks VPS keeps the software cost at zero and bundles the host layer, backups, and 24/7 support into one monthly server bill that you can size to fit the load.

A typical n8n install runs AI agents that call tools, use memory, and step through multi-step tasks on their own. Common adjacent workflows are customer-support chatbots that pull fresh information from a knowledge base before answering, and data pipelines that move records from APIs into a warehouse. On the business side, n8n handles lead-capture and enrichment flows, recruitment screening that ranks candidates from their resumes, and internal-tool syncs that keep your CRM, helpdesk, and project tracker in step. The code node lets you fold custom JavaScript or Python in.

A working production setup wants roughly 2 to 4 CPU cores, 4 to 8 GB of RAM, around 64 GB of SSD storage, and PostgreSQL as the database. The official 2 GB minimum is fine for testing but falls over once you add active workflows. Queue mode with multiple workers, Redis, and a busy PostgreSQL database typically wants 16 GB or more of RAM. The n8n process also needs Node.js between 20.19 and 24.x, and a process manager or container restart policy that keeps it running.

Self-hosted n8n defaults to SQLite, which is fine for testing and small personal setups. For any production use, PostgreSQL is the recommended database because it handles row-level locking, concurrent webhooks, and high execution volumes far better. SQLite uses whole-file locking and starts to degrade after a few hundred executions a day, with a real risk of silently dropping data under heavy load. Switching to PostgreSQL is one of the first changes that long-running self-hosters make once their workflows start to pile up.

Queue mode is a scaling setup where one main n8n instance handles the web UI and event triggering, and a set of separate worker processes runs the workflows. The main instance and the workers communicate through a Redis queue, which lets multiple executions move in parallel across the workers. The pattern is the standard answer for high-volume, webhook-heavy installs that need horizontal scale. It typically wants 16 GB or more of RAM and a VPS with room for the extra worker processes.