Self-hosted AI workflow automation (a Zapier replacement with n8n)
Quick answer
n8n runs multi-step workflows that connect your apps, with AI nodes that call a language model as one step. Self-hosted runs have no per-task fee: cost is the server plus any model API you call. Qdrant can serve as the vector store for retrieval workflows, and Ollama can replace a hosted model for simple tasks. None of n8n publishes an official CPU/RAM figure, so we do not estimate a server size or cost. Size the server for the models and workloads you plan to run.
Last updated · sizing from each app's official docs · prices from Hetzner Cloud (2026-10-22)
Components
- n8n: Workflow engine with integrations and AI agent nodes Official hardware: Not published (measured by us: 607 MiB RAM while working) · License: Sustainable Use License (fair-code) + n8n Enterprise License for .ee files (Source-available)
- Qdrantoptional: Vector store for retrieval (RAG) workflows Official hardware: Not published (measured by us: 566 MiB RAM while working) · License: Apache-2.0 (Permissive)
- Ollamaoptional: Local models for AI steps Official hardware: Not published (measured by us: 1.9 GiB RAM while working) · License: MIT (Permissive)
Combined hardware and monthly cost
| Required components, official minimums | None of the required components publishes an official CPU/RAM figure, so we do not estimate a server size. |
|---|---|
| Smallest server that fits | not estimated |
| Measured RAM in our own test | n8n 2.44.6: 607 MB peak while working; Qdrant 1.20.0: 566 MB peak while working; Ollama 0.40.8: 1.9 GB peak while working. Measured on SelfHostMap test VM (Oct 2026), 2026-10-22; container RAM only, not the whole server; no load test. Model-runner figures are for the small test models named on the app page; larger models need far more RAM. |
| Server sized from our measurement | CPX12 (1 vCPU, 2 GB RAM, 40 GB disk) €11.99/mo Measured in our own test (2026-10-22): n8n 2.44.6 607 MiB working RAM, x1.5 plus 512 MiB for OS and Docker = 1423 MiB, so CPX12 (2 GB). Light use, no load test. Not an official figure. |
| Including optional components | None of the required components publishes an official CPU/RAM figure, so we do not estimate a server size. |
How we got this: we added up the official minimum vCPU and RAM each project publishes (linked on its app page) and took the cheapest Hetzner Cloud Regular Performance (shared AMD vCPU) plan that meets the total. Prices read from www.hetzner.com/cloud/regular-performance on 2026-10-22. "From" monthly price cap shown by Hetzner, excluding VAT. Prices vary by location. Covers the server only; it excludes LLM API usage, backups, extra storage, GPUs and your time. Running everything on one server is our simplifying assumption. Measured figures, where we have them, are on each app page.
Cost vs Zapier
Assumptions and sources
Deployment outline
- Deploy n8n with Docker and a persistent volume (official command on its app page). source
- For hosted models, add an OpenAI Chat Model sub-node (or another provider's) to your AI Agent or chain node. source
- Optional: use local models with the Ollama Chat Model sub-node. source
- Optional: store and search embeddings with the Qdrant Vector Store node. source
This is a summary of the official docs, not a full tutorial. Each app page has the exact official install command.
Gotchas
- n8n: Large payloads, the Code node and manual executions increase memory use and can cause "JavaScript heap out of memory" errors. source
- n8n: The Sustainable Use License allows internal business use only; reselling n8n or hosting it for paying customers is not licensed. source
- Qdrant: The quick-start command runs without authentication and listens on all interfaces. Secure the instance (API key, network) before real use. source
- Ollama: Older NVIDIA cards (compute capability 5.0–6.2) need driver 570 or newer. source
- Ollama: With the Vulkan backend, Ollama needs root or the cap_perfmon capability (sudo setcap cap_perfmon+ep /usr/local/bin/ollama) to read free VRAM; otherwise it schedules models using approximate sizes. source
What this does not replace
- Zapier advertises 9,000+ app integrations; the self-hosted tools ship fewer ready-made connectors, so check that the apps you use are supported.
- Self-hosted runs have no per-task fee, but you run and update the server yourself.
- n8n's Sustainable Use License limits commercial resale; read it before offering n8n to clients.
FAQ
Can I resell automations built on self-hosted n8n?
n8n's Sustainable Use License allows use for your own internal business purposes. Offering n8n itself to others as a paid service is not allowed. Read the license before building client offerings.
How much does a ai workflow automation (zapier replacement) stack cost per month?
None of n8n publishes an official CPU/RAM figure, so we do not estimate a server size or cost. Size the server for the models and workloads you plan to run.
What hardware does n8n need?
None of the required components publishes an official CPU/RAM figure, so we do not estimate a server size..
Can a business use these apps commercially?
n8n: Sustainable Use License (fair-code) + n8n Enterprise License for .ee files (internal business use allowed). Qdrant: Apache-2.0 (internal business use allowed). Ollama: MIT (internal business use allowed). Summaries only, not legal advice.
Is it cheaper than Zapier?
It depends on team size, API usage and how you value your time. Use the calculator on this page; it starts with 10 users and Zapier Team (2,000 tasks/month, up to 25 users) at the price published on 2026-10-22.
Other stacks
- Private ChatGPT for a small business (Open WebUI + LiteLLM)
- Private document Q&A over your company's knowledge (Onyx)
- Self-hosted meeting transcription (Speaches + Open WebUI)
- Self-hosted coding assistant for your dev team (Tabby)
- Self-hosted customer support AI chatbot (Dify)
- Fully offline private AI chat (Ollama + Open WebUI)
Every figure comes from the linked app pages and official docs. Spotted something wrong? Tell us.