How to build an AI agent without code (Lindy, n8n, Make): step-by-step guide
Updated: · VibeStack by Modernitos
This guide contains affiliate links marked as such. If you sign up through them we may earn a commission at no extra cost to you. More info
An AI agent is a language model that, besides answering, can use tools: read your inbox, look something up, update a spreadsheet or send a message, and decide the next step based on the result. You no longer need to code to build one. Here are three approaches, from the most "ready-made" to the most flexible.
Option A: Lindy (a ready-made assistant)
Lindy works like an AI teammate: it lives in Slack, manages your inbox, drafts replies, handles meetings and runs scheduled routines, with 1,500+ integrations and MCP support. According to its pricing page, you start with $50 in free credits (valid 7 days, no card needed). After that, the Team plan costs from $29.99 per user per month for 3,000 credits. By default, actions with outside impact (sending an email, updating a ticket) wait for your approval.
Best for: personal or team assistant tasks, with no building required.
Option B: n8n (maximum control)
n8n has an AI Agent node that you connect to a chat model, memory and tools (other nodes, HTTP requests or MCP servers). It bills per full workflow execution, and the Community Edition can be self-hosted for free. Cloud starts with Starter at €20/month (billed annually), with a trial that needs no card.
- Create a workflow with a trigger (chat, webhook, schedule or a new email).
- Add the AI Agent node and choose a model.
- Attach tools: for example Google Sheets, Gmail, an HTTP request or an MCP client.
- Write the system prompt: role, goals, limits and output format.
- Turn on human approval for sensitive tool calls, then test with real examples.
Option C: Make (visual scenarios with AI)
Make lets you build the agent as a visual scenario: a trigger, AI modules that classify or generate, and then actions across its catalogue of apps. Most actions cost 1 credit, AI features may cost more, and the free plan includes 1,000 credits a month, enough to prototype.
- Create a scenario with a trigger (new email, form, webhook).
- Add an AI module to classify or extract the data.
- Use a router to branch by category: reply, create a task, notify.
- Schedule it and check the run history to tune the prompt.
Start with Make for free → (affiliate link)
Which one?
| You want… | Choose |
|---|---|
| an assistant working today, no building | Lindy |
| full control, self-hosting, high volume | n8n |
| a visual flow and a free plan to experiment | Make |
Other options: Zapier (Zapier Agents) and Gumloop. Comparison: n8n vs Make vs Zapier for AI agents.
Best practices
- Start with one narrow, repetitive task (e.g. classify incoming emails) before you build a "do everything" agent.
- Require human approval for anything that leaves your organisation.
- Give the agent the minimum permissions it needs.
- Log every run and review the errors weekly.
Prices checked on the official pricing pages on 15 October 2026. Plans change often: always confirm on the official site before paying.