How to build an AI agent without code: three routes and the tools for each

The short answer

Without code, there are three ways to build an AI agent: add an “agent” step inside an automation tool (n8n, Make), set up a customer support agent platform (Botpress), or describe the job to an assistant or an agent platform (Lindy, Relevance AI). In all three you supply the same ingredients: instructions, a model, tools, a trigger. The choice depends on where the agent works and on what the tool bills.

If agents are new to you, start with what an AI agent is.

The example: quote requests at a plumbing company

A three-person plumbing company gets quote requests through a web form and by email. The agent has to read each request, check in a spreadsheet that the ZIP code is in the service area, offer a time slot or ask the missing question, and flag emergencies to a person.

Before the tool: the agent’s job description

Five points to write down, whatever the tool.

1. The goal, in one sentence, and what is off limits (“never quote a price”).
2. The trigger: a request comes in through the form or by email.
3. The tools: read the service-area spreadsheet, check the calendar, send an email. The agent can do nothing else.
4. The knowledge: the list of services, business hours, answers to common questions.
5. The guardrails: what needs human approval, and what goes straight to a person (an active leak).

Route 1: an agent inside an automation tool

The agent is one step of a workflow you build: a trigger before it, actions after it. This route gives you the most control over each step, and it is the natural one if you already automate.

n8n. The AI Agent node is included in every Cloud plan and in the Community edition. According to the documentation, it needs a chat model and at least one tool (n8n documentation, “AI Agent” (accessed November 11, 2026)). For the plumber: a webhook trigger tied to the form, then the AI Agent node with three tools. The pricing page also lists human approval of tool calls among the included features (n8n pricing page (accessed November 11, 2026)). Each run of the workflow counts as one execution; models are paid with your own keys or with Gateway credits. More setups are in the n8n workflow examples.

Make. “Make AI Agents” carries a beta label and is listed on every plan. An agent uses 1 credit per operation, plus credits based on tokens when you go through Make’s AI provider. With your own provider, allowed on paid plans, tokens are paid to that provider.

Route 2: a customer support agent platform

The agent lives in a chat window or a messaging app, facing your customers. The platform supplies the channel, the knowledge base and the handoff to a human.

Botpress. A platform for customer support AI agents with a visual builder, billed per conversation. Webchat and WhatsApp are on every plan. According to the vendor, the knowledge base takes websites, PDFs, text and Word files, and spreadsheets. This route fits the plumber if requests come in through the site’s chat instead of email. Tidio follows the same logic with its Lyro agent: it is covered in the AI agent builders comparison.

Route 3: describe the job to an assistant or an agent platform

No workflow to draw here: you describe the task, connect your accounts, and the platform handles the chaining.

Lindy. An autonomous assistant tied to the team’s tools (Slack, email, calendar, meetings), with scheduled routines, billed in credits per user. According to the vendor, actions with an outside effect, such as sending an email, wait for approval (Lindy pricing page (accessed November 11, 2026)).

Relevance AI. A low-code and no-code platform for building agents and agent teams (“Workforces”) connected to apps, billed per action. Your own API keys are accepted on every plan, which avoids “Vendor Credits” for models.

Table scrolls horizontally →

The five tools, row by row

Criterion	n8nAutomation tool	MakeAutomation tool	BotpressAI agent	LindyAI agent	Relevance AIAI agent
Billing unit	executionssource	creditssource	conversationssource	creditssource	actionssource
Free plan	Nosource	1,000 credits per monthsource	Yessource	Nosource	Nosource
AI features	AI Agent node included (Cloud and Community); models paid with your own keys or Gateway creditssource	"Make AI Agents" (beta), every plan: 1 credit per operation, plus token-based credits with Make's AI providersource	AI customer support agents; "AI usage" (model costs) partly included, then $10 credits sold at costsource	Autonomous assistant with scheduled routines and over 40 "skills"; everything is billed in credits from the shared poolsource	Agents and agent teams ("Workforces"); models charged in "Vendor Credits" with no markup, or your own API keyssource
Integrations	"500+ native integrations" (home page); "2,000+ integrations" (integrations page, community nodes included)source	More than 1,000 apps according to the pricing FAQ; "3,500+ apps" on the plan cards of the same pagesource	No total published; integrations named on the site: Intercom, Zendesk, HubSpot, Salesforcesource	"1,500+" according to the documentation; "3,000+" according to the websitesource	"2,000+ Apps" and "1,000+ App Triggers" on all plans, including the Pipedream catalogsource
Vendor	Visit n8n (opens in a new tab)	Visit Make (opens in a new tab)	Visit Botpress (opens in a new tab)	Visit Lindy (opens in a new tab)	Visit Relevance AI (opens in a new tab)

What the plumber’s agent uses on each tool

Assumption: 150 requests a month, and for each one the agent calls three tools (spreadsheet, calendar, email).

• n8n: 150 executions, one per run of the workflow, out of 2,500 on the Starter plan. The model cost comes on top.
• Make: at least 300 credits (one trigger and one agent operation per request). According to the vendor’s help center, the operations of the tools called come on top, along with token-based credits when you use Make’s AI provider (Make Help Center, page on credit usage for AI agents (accessed November 11, 2026)). The Free plan includes 1,000 credits a month.
• Botpress: 150 conversations at most, a conversation being an exchange with at least two messages from the end user in the month. The Plus plan includes 250 a month.
• Relevance AI: 150 × 3, or 450 actions, out of 2,500 a month on the Pro plan. A tool run that fails counts too.
• Lindy: 2 to 250 credits per everyday request according to the vendor, so 300 to 37,500 credits, against a seat that starts at 5,000 credits a month.

These volumes do not compare directly. The guide to billing units explains why, and the cost calculator prices the automation part, AI steps excluded.

Going live, step by step

1. Start with one case: quote requests, not the whole inbox.
2. Run the agent on past requests and read every reply. On n8n, manual executions started from the editor do not count toward the quota.
3. Keep human approval on anything that commits the business: a message to the customer, an appointment placed on the calendar.
4. Watch usage the first week: that is when you see how many tools the agent really calls per request.
5. Then widen the scope, one case at a time.

Building an AI agent for free: what the plan grids say

• n8n: no permanent free plan on n8n Cloud. Free use exists only when self-hosting, with the Community edition, which includes the AI Agent node. See is n8n free?
• Make: the Free plan gives access to agents, but connecting a custom AI provider is reserved for paid plans.
• Botpress: the pricing page says “100 conversations” for the Free plan, with no period stated, and 3 AI agents.
• Lindy: no permanent free plan; free credits valid for 7 days.
• Relevance AI: the Free plan is retired.

Which route for which situation

• You already automate and want control over every step: route 1, with n8n or Make.
• The agent answers your customers in a chat: route 2, with Botpress.
• You want to delegate office tasks without drawing a workflow: route 3, with Lindy or Relevance AI.
• The agent has to use tools outside the platform: check support for the MCP protocol.