n8n AI Nodes: Automate Your Business with Self-Hosted AI

n8n already connects your applications with automated workflows. With AI nodes, it connects your applications to an AI brain running on your own server. A customer email arrives, the AI reads it, drafts a reply, and saves it as a draft for human review. A support ticket comes in, the AI categorises it and assigns it to the right team. A new lead appears in your CRM, and the AI researches the company and writes a briefing note. All of this runs on your infrastructure with models you control.

What n8n AI Nodes Can Do

n8n includes native nodes for Ollama and OpenAI-compatible APIs. Point the node at your self-hosted Ollama instance, and every n8n workflow can call on an AI model for reasoning, classification, summarisation, or generation. The AI node is just another step in the workflow, between a trigger (new email) and an action (create draft reply).

Practical workflows you can build today:

• Email triage: Monitor a shared inbox. When a new email arrives, the AI reads it, classifies it (sales inquiry, support request, billing question, or spam), and routes it to the appropriate team channel in Mattermost with a summary.
• CRM enrichment: When a new lead is created in EspoCRM, the AI researches the company from publicly available information and writes a summary briefing note attached to the lead record.
• Support response drafts: When a support ticket arrives in FreeScout, the AI reads the customer's question, searches your knowledge base in Outline for relevant articles, and drafts a reply that a human agent reviews and sends.
• Meeting note summarisation: Paste meeting notes into a Nextcloud document. n8n detects the new document, sends it to the AI, and the AI produces a structured summary with action items, decisions made, and follow-ups needed. The summary is posted to the relevant Mattermost channel.
• Invoice data extraction: An invoice PDF arrives in a Paperless-ngx consume folder. n8n detects it, sends it to the AI for data extraction (vendor, amount, date, invoice number), and creates an entry in your accounting system.

Setting Up the Ollama Node in n8n

n8n includes an Ollama node by default. No plugin needed. To configure it:

1. In n8n, add an Ollama node to your workflow
2. Set the Base URL to your Ollama server (http://ollama-server:11434)
3. Select the model (mistral:7b, llama3.1:8b, etc.)
4. Write the prompt using n8n's expression syntax to inject data from previous workflow steps

The prompt can reference any data from earlier in the workflow. For example: Summarise the following email and suggest a reply. Sender: {{ $json.from }}. Subject: {{ $json.subject }}. Body: {{ $json.body }}. The AI processes the prompt and returns a response that flows to the next step in the workflow.

Keeping AI Costs at Zero

With Ollama running on your own hardware, every AI-powered workflow step costs nothing beyond the electricity to run the server. There are no API call charges, no token limits, and no per-use pricing. A workflow that runs 500 times per day costs the same as one that runs 5 times per day. This changes the economics of AI automation. You can afford to run AI on every email, every ticket, and every lead, because the marginal cost of each AI call is zero.

Contrast this with OpenAI's API pricing. Processing 500 emails per day through GPT-4 API would cost hundreds of dollars per month. Through your own Ollama instance, it costs nothing. The quality difference between GPT-4 and Mistral 7B for email classification and draft replies is small. The cost difference is infinite.

VPS1 deploys n8n with AI nodes connected to your self-hosted models. We build your first AI-powered workflows and train your team to build more. Your business processes get smarter. Your AI costs stay at zero.