Connect a Local LLM to n8n: Step-by-Step Tutorial (2026)
Run Ollama or LM Studio inside n8n for free, private data extraction. Full setup, working JSON, and the gotchas nobody warns you about.
n8n tutorials
Step-by-step n8n tutorials written by people who run n8n in production, not by people demoing the welcome screen. Every guide ships with a JSON export, the prompts we actually use, and the breakages we hit so you skip them.
Key takeaways
- Self-host n8n on a $10 VPS once you pass 5,000 workflow runs a month. The cloud tier math stops working fast.
- Always import a credentials secret rather than pasting keys into HTTP nodes. Future-you debugging at 11 pm will be grateful.
- Wrap any AI node in a Try/Catch with a fallback branch. LLMs return malformed JSON often enough to matter.
- Pin n8n version in docker-compose. Auto-updates have broken our workflows three times in 2 years.
Frequently asked questions about this category
Is n8n a good first workflow tool for someone coming from Zapier?
Yes, if you can read JSON and you ever expect to need more than 1,000 runs a month. The learning curve is steeper for one week, then n8n is faster to build in than Zapier forever after. The cost difference at any volume past a few thousand runs is dramatic.
Should I run n8n self-hosted or use n8n Cloud?
Self-host on a $10/month Hetzner or DigitalOcean VPS the moment you hit 5,000 runs a month. Below that, n8n Cloud is the right call for the time it saves on setup. The break-even has shifted in 2026 because cloud prices went up while VPS prices stayed flat.
Can n8n connect to local LLMs like Ollama?
Yes. n8n has a dedicated Ollama node and an AI Agent node that points at any OpenAI-compatible endpoint, which includes LM Studio and llama.cpp. Latency is 150 ms to 2 s depending on model size. See our local LLM + n8n tutorial for a full setup.
How do I debug a failing n8n workflow?
Three steps. Open the execution log and click the failed node to see the actual input payload. Add a Set node before the failure to inspect intermediate data. If the bug is in an HTTP request, switch on "Always Output Data" and capture the raw response into Postgres for replay.
What does n8n cost to run at 100,000 webhook events a month?
Self-hosted n8n on a $20/month VPS handles 100k webhook events comfortably, total cost ~$20 plus whatever AI API spend the workflows incur. The same volume on Zapier (Professional tier) would cost $599/month. The cost gap is why most teams that scale graduate to n8n.
Automating Social Media Content with LLMs and n8n
Build a complete social media content pipeline using n8n and LLMs that generates posts, creates images, schedules publishing, and tracks engagement — fully automated.
How to Build a Self-Correcting LLM Agent Workflow in n8n
Learn how to build an AI agent workflow in n8n that checks its own output, catches mistakes, and retries until the result is correct — no coding required.
How to Connect Local LLMs to Automation Workflows (Private AI Guide)
Learn how to connect local LLMs to automation workflows for private, cost-free AI processing. This step-by-step guide covers Ollama, LM Studio, and LocalAI with n8n, Zapier, and other automation platforms.
Build a Self-Correcting RAG Pipeline with n8n and a Local LLM
Combine retrieval-augmented generation with self-correcting agent loops in n8n. Use Ollama for private inference, a vector store for grounded answers, and a critique step that detects bad responses and retries with better context.
A Beginner's Guide to n8n: Your First AI Workflow in 30 Minutes
Install n8n, build your first AI-powered workflow, and trigger it with a webhook — all in 30 minutes. No prior automation experience required.