7 Free n8n AI Agent Templates You Can Import in 5 Minutes

Last updated October 2026 · ~9 min read · by StackHonest

Short answer: the official n8n template library has over 12,000 community workflows, every one free to import, and the seven below are the AI agent templates I'd actually point a beginner at — a first-agent tutorial, two Telegram assistants, a RAG document chatbot, database chat, Gmail auto-labelling, and Sheets-via-chat. "Import in 5 minutes" is honest for the import itself; budget another 15–30 minutes per template for the part nobody mentions: wiring your own API keys and OAuth credentials. The software side is $0 on self-hosted n8n (or Cloud from $20/mo billed annually, ~$24 month-to-month, 2,500 executions — verified at n8n.io/pricing, October 1, 2026); the LLM API usage is the real running cost, and I break it down per template below.

What Makes an AI Agent Template Actually Usable

The Template availability currently offers no download or email signup. The library is community-submitted — some templates are polished teaching tools, others are one person's half-configured weekend project exported as-is. Before you burn an evening, check four things:

• Sticky notes on the canvas. Good templates annotate every section: what to configure, which credentials to create, what each branch does. If the canvas is bare nodes with no notes, expect archaeology.
• Standard nodes over exotic ones. Templates built on core nodes (AI Agent, HTTP Request, Gmail, Telegram, Google Sheets) survive n8n version updates. Ones that lean on obscure community nodes break the moment you're a version apart.
• Credential count. Every external service is a credential you must create — an OpenAI or Gemini key, a Telegram bot token, Google OAuth consent. Three credentials is a 20-minute setup; seven is an afternoon.
• A recent update date and real usage numbers. The library shows both. AI nodes evolved fast through 2025–2026; a template last touched two years ago will import, but the model names it references may not exist anymore.

All seven picks below pass that filter: they're from the library's AI category, built mostly on standard nodes, and annotated well enough that I'd hand them to someone on week one. (New to n8n entirely? Read the n8n review for beginners first — it covers the three gotchas that make people quit on day two.)

The 7 Templates (Import Links + What Each Does)

All links go to the official n8n library. On each page, hit "Use for free" to push the workflow into n8n Cloud or copy the JSON for a self-hosted instance (paste it straight onto a blank canvas).

1. Build Your First AI Agent — Lucas Peyrin

Import link. Start here even if another template is your real goal. It's an interactive tutorial disguised as a workflow: a chat agent with a system prompt, memory, and tools, with sticky notes explaining every piece as you click through. The mental model it teaches — agent node + model + memory + tools — is the same architecture every other template on this list uses. What breaks: almost nothing, which is why it's first; it's built to run with minimal setup.

2. Angie, Personal AI Assistant with Telegram (Voice and Text) — Derek Cheung

Import link. The classic Telegram assistant: message or voice-note a bot, it transcribes, reasons, and answers — summarizing emails, checking your calendar and tasks. What breaks: Google OAuth. Gmail/Calendar credentials mean creating a Google Cloud project, enabling APIs, and adding yourself as a test user — the single most common place beginners stall. The Telegram bot token (via @BotFather) takes two minutes by comparison.

3. Personal Life Manager with Telegram, Google Services & Voice AI — Derek Cheung

Import link. Angie's bigger sibling: multiple Google services wired into one voice-enabled agent that manages your actual life admin. More capable, more credentials (~6+). What breaks: same OAuth gauntlet as #2, multiplied. Do Angie first, then upgrade — the credentials you created carry over.

4. RAG Chatbot for Company Documents (Google Drive + Gemini) — Mihai Farcas

Import link. The "chat with your documents" pattern done properly: it watches a Drive folder, chunks and embeds new documents into a vector store, and answers questions grounded in your files instead of the model's imagination. This is the template I'd show a client who asks what AI agents are actually for. What breaks: the vector store. You'll stand up a Postgres/pgvector or similar database and swap in its credentials — the template can't ship one. Budget the full 30 minutes.

5. Chat with a Database Using AI — David Roberts (n8n team)

Import link. Ask questions in English, get answers from your SQL database — the agent writes and runs the queries. By an n8n team member, so it tracks the current node architecture. What breaks: nothing in the template; the risk is in what you point it at. Give it a read-only database user. An LLM that can write arbitrary SQL against production with write permissions is a horror story you only get to tell once.

6. Basic Automatic Gmail Labelling with OpenAI — Martijn Kerver

Import link. The quiet workhorse of the list: new mail arrives, the model reads it, applies the right Gmail label. No chat interface, no demo sparkle — it just runs every day and keeps your inbox sorted. This is the template most likely to still be running in six months. What breaks: label mismatches — the labels in the prompt must exist in your Gmail, and you'll want to edit the category list to match how you actually sort mail. Classification is cheap-model work; see the cost section.

7. Talk to Your Google Sheets Using ChatGPT — Robert Breen

Import link. Chat with a spreadsheet: "what did I spend on ads in March?" against a Sheet instead of eyeballing rows. If your business runs on spreadsheets (most one-person businesses do), this is the fastest useful agent here. What breaks: messy sheets. The agent inherits your tab names and column headers — if the Sheet is chaos, the answers will be too. Clean headers first, then import.

Hidden Costs: API Keys and LLM Usage

"Free template" means the workflow JSON costs nothing. Running an AI agent has two real costs, and nobody puts them on the template page:

1. The n8n instance. Self-hosted Community Edition is free software — unlimited workflows, unlimited executions — on a $5–7/mo VPS (my cheap self-hosting walkthrough covers the whole setup). n8n Cloud Starter is $20/mo billed annually (~$24 month-to-month) for 2,500 executions; Pro is $50/mo annual for 10,000 (verified at n8n.io/pricing, October 1, 2026). One caution for agent workflows on Cloud: a chatty Telegram assistant fires an execution per message, so 2,500/month is roughly 80 messages a day — fine for personal use, tight if your whole family adopts the bot. Full math in n8n Cloud vs self-hosted cost.

2. The LLM API. Every template above calls a model with your key, billed per token. Current list prices (verified October 1, 2026 against platform.openai.com/docs/pricing and docs.anthropic.com):

What that means in practice: a personal assistant handling 300 messages a month at ~2,000 tokens per round-trip is well under $1 on a budget model and a few dollars on a mid-tier one. The Gmail labeller classifying 1,500 emails a month on gpt-6-luna costs pennies. The RAG chatbot is the one to watch — embedding a large document folder is a one-time spike, and each answer stuffs retrieved chunks into the prompt, so input tokens run 5–10x a plain chat message. Set a spending cap in your API dashboard before you hand any bot to other people; that's not paranoia, it's the difference between a $3 month and a surprise $40 one.

How to Customize a Template Without Breaking It

Every template is someone else's defaults. The order of operations that has saved me the most grief:

1. Run it stock first. Credentials in, zero other edits, one successful execution. If it breaks after that, you broke it — which is useful information. If it never ran stock, you're debugging two problems at once.
2. Change the system prompt before the structure. 80% of "customizing an agent" is prompt text: its name, tone, your categories, your rules. That's editable with zero risk to the wiring.
3. One node at a time, pin your test data. Swap the model, test. Add a tool, test. n8n lets you pin a node's output so you can re-run downstream changes against the same input without re-triggering Telegram or Gmail every time — use it.
4. Duplicate before surgery. Structural changes (new branches, different memory, extra tools) happen on a copy. The original stays importable-state forever, which beats re-importing and re-wiring credentials from scratch.
5. Add error handling before you depend on it. Community templates almost never ship with it — when the API times out at 3am, the default is silent failure. Fifteen minutes with an Error Trigger workflow fixes that; exact settings in n8n error handling and retry workflows.

Templates sit at a useful midpoint: more concrete than building from a blank canvas, less magic than the "describe an agent and AI builds it" demos making the rounds (covered honestly in Claude vs ChatGPT for automation workflows). You'll learn more fixing someone's working agent than generating a fresh one you don't understand. The Template availability currently offers no download or email signup. Related reading: best AI automation tools.

FAQ

Are n8n AI agent templates really free?

Yes — all 12,000+ workflows in the official library are free to import. The costs are the n8n instance (free self-hosted; Cloud from $20/mo billed annually) and your own LLM API keys, billed per token — typically under $5/month for personal-assistant use.

How do I import an n8n template?

Click "Use for free" on the template page — it pushes into n8n Cloud directly or copies the JSON for self-hosted (paste onto a blank canvas). Import takes a minute; connecting your API keys and OAuth credentials is the real setup time.

What does it cost to run an n8n AI agent?

Instance + tokens. Self-hosted n8n is free on a $5–7/mo VPS; Cloud Starter is $20/mo billed annually for 2,500 executions. On the API side (Oct 2026 prices): gpt-6-luna at $0.10/$0.50 per million tokens handles classification for pennies; Claude Haiku 4.5 at $1/$5 and Sonnet 5.5 at $2/$10 cover assistant and RAG work for a few dollars a month at personal volume.