n8n templates

tl;dr: 12,237 templates at n8n.io/workflows, with 8,454 AI workflows at /workflows/categories/ai/. Treat every template as a starting point: read the nodes, swap credentials for yours, test before you put it in front of real data.

What “n8n templates” actually are

Each template is a JSON-serialized workflow: a graph of nodes (triggers, integrations, AI steps, code, HTTP) plus credentials and parameter defaults. Open one in the editor, hit Use this workflow, and the JSON gets dropped into your instance with empty credential placeholders — you fill in API keys and connect accounts before it runs.

Counts from the n8n.io/workflows hub page (fetched 2026-09-10):

Categories that aren’t AI have lower four-digit counts but weren’t enumerated on the hub page this session.

Browse by what people actually search

The n8n.io/workflows hub lets you filter by integration (Google Sheets, OpenAI, Telegram, Gmail, MySQL, Postgres, Discord, Google Drive, Slack, Notion, Outlook) or by use case. The AI category page surfaces the same filters plus the Verified Creators shelf — accounts whose templates get a badge after review.

Verified creators visible on this fetch (counts from the hub):

n8n runs a Creators program: submit templates, get featured, earn via the affiliates program.

How to import a template

In the editor:

1. Go to the template page on n8n.io/workflows.
2. Click Use this workflow (cloud) or Use workflow (self-host) — the JSON drops into your instance.
3. Open each node flagged with a credentials warning. Add the API key or OAuth for that service.
4. Test the workflow with a tiny input — chat messages, one row, one file. Confirm each node fires.
5. Activate.

If you want the JSON by hand (e.g. to commit it to your own repo), the template page exposes a Copy workflow JSON button. Paste into n8n import:workflow --input=workflow.json (self-host CLI) or the editor’s Import from clipboard option. Full import docs: docs.n8n.io.

For machine-readable discovery, the llms.txt index lists templates and categories as structured text.

Common AI templates people actually fork

The AI category leans heavily on the same handful of patterns. From the AI category page this session:

• Chat-with-data RAG — webhook/chat trigger → vector store retrieve → OpenAI/Claude/Ollama → Slack/email response. See n8n.io/ai-agents and n8n.io/rag for the pattern.
• Telegram / Slack / Discord bot — chat trigger → agent node → tool calls (HTTP, Postgres, calendar) → reply.
• Email triage — Gmail/Outlook trigger → classify with an LLM → branch by intent → Slack alert + auto-draft.
• Lead enrichment — webhook from a form → enrich with Clearbit-style HTTP call → push to HubSpot/Salesforce via node.
• Scheduled summarizer — cron → fetch RSS/API → summarize → post to Notion or email digest.
• IT ops alert — webhook from monitoring → agent decides severity → PagerDuty/Slack node.
• Security ops — SIEM webhook → LLM summarization → ticket system. See n8n.io/secops.

Gotchas (from the fetch)

• Community-submitted. The hub copy is explicit: templates are starting points, not production copy-paste. Read every node, swap credentials, test.
• Credentials are not bundled. Every template that talks to a third-party service has empty credential placeholders. You add your own API key or OAuth on first run. Do not paste someone else’s credentials.
• Fair-code license. Sustainable Use License lets you self-host and modify, but commercial use above the threshold needs a paid license. Templates inherit the license of the underlying n8n, not a custom one.
• n8n’s own cloud is in Frankfurt, EU (security). Templates on the public hub are not stored in your region; the workflow JSON you import is what runs in your instance.
• Pricing is per execution, not per task. The AI category FAQ claims the pro plan starts around $50/mo where comparable platforms charge $500+/mo for 100k tasks. Check the current pricing page — numbers change.
• SOC2 compliant (security page), RBAC, encrypted secret storage. Enterprise adds SAML/LDAP.
• Verified vs not. A “verified” badge means n8n reviewed the template; absence of the badge means a community submitter. Both work, neither is a guarantee.
• MCP servers can drive workflows. n8n exposes /mcp-server/http for Claude Code, Codex, and Claude Desktop — templates can also be created/edited from inside those clients via n8n.io/mcp.

Publishing your own

The same hub you read from is where you submit back to. creators.n8n.io/hub is the submission form; review typically takes a few days. Featured templates get a creator page (e.g. derruttie, portio) and earn via the affiliate program.

Related: n8n for the runtime, Claude Code and Claude CLI for the MCP clients that drive templates, best local LLM if your AI template calls an Ollama or llama.cpp endpoint.