Build Your First No-Code AI Agent in One Day

A one-day plan to build an email-drafting AI Agent in n8n: design on paper, connect tools and memory, then test with 20 real emails.

You have read several articles about AI Agents. You understand the loop, the tools and the risks, but you have never built one yourself. This article is a one-day plan for people who do not write code to get a first agent doing real work, using n8n as the example. If n8n is new to you, read What is n8n first; it will help a lot.

What is a no-code agent

It means assembling an AI Agent from ready-made blocks on screen instead of writing code. In n8n the core is a node called AI Agent, connected to several sub-nodes: a chat model as the brain, tools as the hands, and memory for ongoing conversations. n8n's documentation lists key options for this node such as System Message, Max Iterations (default 10), Return Intermediate Steps and Require Specific Output Format, which map directly onto the ideas in The Agent Loop.

Example task: an agent that drafts replies to product enquiries

A small office supplies shop gets dozens of emails a day asking about prices and stock. The agent's goal is to read the email, look up price and stock in the product sheet, and draft a reply for a person to check before sending. This task is a good choice because:

• It involves real decisions along the way, such as which products to look up and how to reply when something is out of stock
• It needs only 2 tools: read the sheet and create an email draft
• It is low risk, because nothing is sent automatically
• It is easy to measure: compare the agent's drafts with replies the team actually sent

The one-day plan

Early morning: design on paper (1 to 2 hours)

Write down 4 things: the agent's goal, what a good draft looks like, the tools it needs, and what it must not do, such as offering discounts or promising delivery dates. Then collect 20 real emails you have received to use as tests.

Late morning: build the skeleton

1. Create a new workflow with a trigger that receives incoming email, or use a chat trigger for testing first
2. Add an AI Agent node and connect a chat model you have an account for
3. Write the System Message: the role, the steps, what is not allowed, and an instruction to ask back or mark "needs human review" when information is missing

Afternoon: add hands and memory

1. Connect a tool that reads the product sheet, with a clear description of what it looks up and with what input
2. Connect a tool that creates an email draft, not one that sends email
3. If you need ongoing conversation, attach a memory sub-node as n8n's documentation recommends
4. If one day you want it to send by itself, turn on human-in-the-loop for that tool. n8n supports approvals through channels such as Slack, Microsoft Teams or Gmail

Evening: test and refine

Run all 20 test emails. Turn on Return Intermediate Steps to see which tools the agent called and with what values, then note which cases went wrong and why. Fix one thing at a time and run again. For how to log and classify errors, see Why Agents Fail.

Watch out for

• Credentials and permissions: connect limited-permission accounts and use a copy of the sheet with only the columns the agent needs
• Customer data: customer emails are personal data. Check company policy and PDPA requirements, or consult a professional, before going live
• Cost: every email means several model calls. Check your provider's pricing and set Max Iterations sensibly
• Do not skip ahead to auto-sending: stay with drafts until test results are stable

Your first agent does not need to be brilliant. It just needs to do one real job and let you see every step of its thinking.

What to do after day one

1. Keep using drafts for another 1 to 2 weeks and note how often people have to edit them
2. Turn emails the agent got wrong into test tasks, following the approach in Evals
3. Add new tools one at a time, deciding which ones need human approval
4. Write a short note on how much time it saves, to use as evidence of your work

If you prefer to learn step by step, start with the n8n Basics course and continue with n8n AI on Plants.work Learn. Your first agent is also something you can showcase on your resume with Resume Lab and talk about in interviews.