AI Support Agent

After this module you can build a customer support chatbot that looks up real data, takes actions, remembers the conversation, and knows when to hand off to a human.

What it does

A hosted chat page (n8n Chat Trigger) talks to an AI Agent for a demo coffee shop, Brewlab Coffee Co. The agent follows store policies from its system prompt, remembers the last 10 messages per session, and uses two tools: lookup_order (status, carrier, tracking, ETA) and create_support_ticket (hand-off to humans).

Flow

Support Chat → Support Agent
                 ├─ model:  OpenAI Chat Model (gpt-5-mini)
                 ├─ memory: Conversation Memory (10 messages per session)
                 └─ tools:  lookup_order, create_support_ticket

Verification status

• Hosted chat page: tested on n8n 2.43.6. Messages reach the agent and stop at "Error in sub-node OpenAI Chat Model" without a credential, as expected.
• Agent behavior (tools, memory, hand-off, staying on topic): needs an OpenAI credential. WITH_CREDS=1 ./test/test.sh runs a 5-turn conversation.

Files

File	Tier	Purpose
lesson.md	Free	Step-by-step build
workflow/starter.json	Free	Chat trigger + order tool only
workflow/learner.json	Learner	Finished agent
workflow/business.json	Business	Business workflow (see setup guide)
workflow/business-create-ticket.json	Business	Ticket sub-workflow
assets/embed.html	Business	Website chat widget
setup-guide.md	Business	Credentials + configuration
exercises.md	Learner	Practice tasks
solutions.md	Learner	Exercise solutions
notes.md	Private	Learning log
sell/	Private	Offer, video script, blog post