Chapter 05 · AI Agents AI Agents Track No-code

A fleet of six QA agents, built in n8n without writing a line of code.

Drag-and-drop AI agents that answer QA questions, file Jira bugs from plain English, and turn PRDs into test plans, Jira subtasks, and Excel test-case exports. Runs on Groq cloud or a fully local Ollama model.

01 What students build

Six progressively harder n8n workflows, all sharing one pattern: a Chat Trigger feeds a LangChain AI Agent node wired to an LLM, buffer-window memory, and tools. Agents 01 and 02 are QA expert chatbots (cloud Groq vs local Ollama). Agents 03 to 06 add Jira tool-use: create bugs from conversation, read a PRD ticket and write a test plan, generate test cases as Jira subtasks, and export test cases straight to Excel.

Cloud or Local LLM

The same QA expert agent runs on Groq's gpt-oss-120b or on a private gemma3:1b via Ollama, so students feel the cost, privacy, and quality trade-off first-hand.

Jira Tool-Use

Agents call the Jira Software node as a tool, with AI-powered field mapping ($fromAI) extracting summary, description, and issue type from chat.

PRD to Artifacts

The top agents read a PRD ticket, design test cases, then either create them as linked Jira subtasks or convert them to rows and export a downloadable .xlsx.

02 Architecture

Chat Triggermessage / webhook
──▶
AI AgentLangChain node
──▶
Jira Toolread / create issue
──▶
Outputreply · ticket · .xlsx
Groq LLMgpt-oss-120b
Ollamagemma3:1b local
Buffer Memorylast 10 messages
Spreadsheetrows → Excel
six workflows · one pattern: trigger → agent ← llm + memory + tools

03 Live demo (simulated with sample data)

This is a UI walkthrough with pre-baked sample data so you can feel the workflows without a running instance. In the course you import the six JSON workflows into a real n8n at localhost:5678 with Groq, Ollama, and Jira Cloud credentials.

n8n · localhost:5678 demo · sample data

04 Inputs & outputs

What goes in

  • Chat messages in n8n's built-in chat panel (or webhook calls)
  • Plain-English bug descriptions
  • Jira ticket ids pointing at PRDs and stories (VWO-48, VWO-105)
  • Credentials: Groq API key, Jira Cloud token, local Ollama URL

What comes out

  • Grounded QA and automation answers with conversation memory
  • Jira Bug tickets created automatically with mapped fields
  • Comprehensive test plans generated from PRD tickets
  • Test cases as linked Jira subtasks
  • Downloadable .xlsx / .csv test-case sheets

05 How it's built

Install n8n (npm, Docker, or n8n.io cloud) and open the visual editor at localhost:5678.
Import the six workflow JSON files from the chapter and attach credentials: Groq, Ollama, Jira Cloud.
Build Agent 01: Chat Trigger, AI Agent node with a Senior QA Expert system prompt, Groq chat model, buffer-window memory (10 messages).
Swap the LLM for a local Ollama gemma3:1b in Agent 02 to go fully private and offline.
Add the Jira Software node as an agent tool: create bugs from chat (Agent 03), then read PRD tickets to draft test plans (Agent 04).
Chain generation with action: create test cases as Jira subtasks (Agent 05) and export structured rows to Excel (Agent 06).

06 Tech stack

n8nLangChain Agent nodeGroq gpt-oss-120bOllama gemma3:1bJira Software CloudBuffer Window MemoryChat / Webhook triggersSpreadsheet + Excel export$fromAI field mapping