---
title: "How to build an AI agent (2026): 6 builders graded, with steps"
description: "How to build an AI agent in 2026: Make AI Agents is the top pick of 6 builders graded 0 to 3, ahead of Zapier, n8n, Lindy, Copilot Studio and the OpenAI SDK."
canonical: "https://thenotetakers.com/how-to-build-an-ai-agent"
last_updated: "2026-11-02"
---


# How to build an AI agent (2026): 6 builders graded, with steps

**To build an AI agent with no code, start with Make AI Agents. No builder reaches grade 3 on this page's depth scale, and Make is grade 2 (modules, scenarios and MCP tools) on its Free plan at $0, with 1,000 credits a month. The six steps below apply to every builder on this page.**

What changed: Nov 2, 2026: first edition; 6 builders graded 0 to 3, none at grade 3, Make AI Agents the pick.

## How to build an AI agent: Make AI Agents is the top pick

Make AI Agents is grade 2 for no-code agents on app workflows, on Free at $0 (1,000 credits a month; [Make pricing](https://www.make.com/en/pricing), checked Nov 2026).

No builder reaches grade 3. Grade 3 needs memory the agent writes itself and keeps across runs, and none of the six documents that without a limit. Five builders are grade 2 and are shelved by plan, then price: Make and Zapier Agents on Free at $0, n8n on Starter at 20€ a month billed annually, Lindy on Team at $49.99 a month, and Copilot Studio at $200 a month per capacity pack. The OpenAI Agents SDK is grade 0 as shipped.

Make and Zapier Agents tie on grade, plan and price, so they are listed alphabetically. Pick Make if you already work in Make, since its agents are "built, run, and debugged directly inside the same Make canvas you already use"; pick Zapier Agents if you need its 9,000+ apps or want to start an agent from Zap workflows or MCP.

## How the 6 builders are graded

Each builder gets a depth grade from 0 to 3. The grade counts only what the vendor's own docs say the product does as shipped.

| Grade | Meaning |
| --- | --- |
| 0 | Chat only: no prebuilt app actions or triggers in the product |
| 1 | Triggers or [webhooks](/send-meeting-data-with-webhooks) start the agent |
| 2 | Native app actions: the agent acts in your apps through the builder's own connectors |
| 3 | Native app actions, plus tools or MCP, plus memory: the agent writes it itself and keeps it across runs. Files or knowledge you upload do not count, and neither does memory that lasts only one run. Memory in preview for one harness only, or memory that expires, counts as Partly and stops the grade at 2 |

Code you write yourself is printed beside the grade, as in "0 (2 in code)", and does not move the entry.

"Plan required" is the lowest vendor-hosted plan that keeps the grade with no time limit. A free self-hosted edition, or free credits that expire, are printed beside the plan and do not move the entry. Where a vendor page does not state something, the entry says "Not documented".

## AI agent builders compared

Shelved by depth grade, then plan required (Free, then entry paid plan, then higher, then sales-only), then the lowest list price for that plan. Ties after all three are alphabetical. Last checked: Nov 2, 2026.

| Rank | Builder | Depth grade | Plan required | Lowest list price | Code needed | Last checked |
| --- | --- | --- | --- | --- | --- | --- |
| 1 | Make AI Agents | 2 | Free | $0, 1,000 credits a month | No-code | Nov 2, 2026 |
| 2 | Zapier Agents | 2 | Free | $0, 400 activities a month | Not documented | Nov 2, 2026 |
| 3 | n8n | 2 | Starter (free Community edition if self-hosted) | 20€ a month, billed annually | Not documented | Nov 2, 2026 |
| 4 | Lindy | 2 | Team (Free credits last 7 days) | $49.99 a month, 5,000 credits | No, set up from a sentence | Nov 2, 2026 |
| 5 | Microsoft Copilot Studio | 2 | Capacity pack (free trial offered) | $200 a month per pack of 25,000 Copilot Credits | Low-code | Nov 2, 2026 |
| 6 | OpenAI Agents SDK | 0 (2 in code) | None, open source (MIT license) | $0, MIT license | Python | Nov 2, 2026 |

## How to build an AI agent in 6 steps

1. **Pick one job.** Choose one task you repeat, with a clear finish. Lindy puts it this way: "Whatever you do on repeat, hand it to Lindy on a schedule."
2. **Choose a builder.** Use the table above. Grade tells you how far the agent can reach on its own. Plan required tells you what you pay to keep that reach.
3. **Connect your apps.** Make lists 3,000+ native apps, Zapier 9,000+ apps, n8n 2,000+ integrations and Lindy 1,500+ integrations. Copilot Studio uses prebuilt connector tools and triggers.
4. **Add instructions and tools.** In Make, tools "can include modules, scenarios, and MCP tools". n8n's AI Agent node states: "You must connect at least one tool". The OpenAI Agents SDK defines agents as "LLMs equipped with instructions and tools".
5. **Test before it runs on its own.** On Zapier Free, tests count toward the monthly activity quota. Copilot Studio lets you describe an agent in plain language, "then test it before you publish."
6. **Add guardrails.** In Lindy, "actions that write or send can wait for your approval." Agents SDK guardrails validate agent inputs and outputs. Copilot Studio event triggers use the agent maker's credentials, so scope that account's access with care.

## 1. Make AI Agents

Make AI Agents is grade 2 for no-code agents on app workflows, on Free at $0 (1,000 credits a month; make.com/en/pricing, checked Nov 2026).

- **Grade:** 2. Tools include modules, scenarios and MCP tools. Context files give the agent extra information. Make calls them long-term memory, but its docs do not say the agent writes them itself, so the grade stops at 2.
- **Plan required:** Free. Make says AI Agents are "available on all plans using Make's AI Provider".
- **Price:** $0 on Free, with 1,000 credits a month. Paid plan prices: Not documented.
- **Connects to:** 3,000+ native apps (the same page also says 3,500+ pre-built integrations).
- **Tools and MCP:** You can connect an agent to MCP servers for more tools.
- **Memory:** Context files stored in the agent's "long-term memory" give it extra information. Memory the agent writes itself across runs: Not documented.
- **Best for:** teams that already run Make scenarios.
- **Limitation:** most actions use 1 credit, but some features that use Make's AI Provider may use more. Custom AI provider connections need a paid plan.
- **Last checked:** Nov 2, 2026.

## 2. Zapier Agents

Zapier Agents is grade 2 for app actions across 9,000+ apps, on Free at $0 (400 activities a month; zapier.com/pricing, checked Nov 2026).

- **Grade:** 2. Agents act across 9,000+ apps and draw on company knowledge you add. Agent memory and an MCP client inside Agents: Not documented.
- **Plan required:** Free.
- **Price:** Free, 400 activities a month and 10 per run. Pro starts at about $33.33 a month billed annually, with 1,500 activities a month and 40 per run.
- **Connects to:** 9,000+ apps. Triggers include On Demand, Schedule by Zapier, Zap workflows and MCP, and app triggers such as Gmail, Google Sheets or Slack. Teams that already run [Zapier workflows](/meeting-notes-zapier-workflows) can start an agent from a Zap.
- **Tools and MCP:** agents can be triggered via Zap workflows and MCP; an MCP client inside Agents is Not documented.
- **Memory:** Not documented. Company knowledge you add is reference content, not memory.
- **Best for:** the widest app catalog.
- **Limitation:** Zapier is migrating Agents to AI by Zapier and has not set a date to turn Agents off.
- **Last checked:** Nov 2, 2026.

## 3. n8n

n8n is grade 2 for agents you can self-host, from Starter at 20€ a month billed annually (n8n.io/pricing, checked Nov 2026).

- **Grade:** 2. In the AI Agent node, "the agent decides which tools to call to complete a task", and the MCP Client Tool node uses tools exposed by an external MCP server. Simple Memory "stores a customizable length of chat history for the current session"; memory kept across runs is Not documented, so the grade stops at 2.
- **Plan required:** Starter. Without a license key, n8n runs as the free Community edition when you host it yourself.
- **Price:** Starter 20€ a month billed annually, with 2.5K workflow executions. Pro is 50€ a month billed annually, with 10K.
- **Connects to:** 2,000+ integrations for apps, services and APIs.
- **Tools and MCP:** the AI Agent node takes a chat model and one or more tools; the MCP Client Tool node connects external MCP tools.
- **Memory:** Simple Memory keeps chat history for the current session. Across runs: Not documented.
- **Best for:** technical teams that want to self-host.
- **Limitation:** the AI Agent node needs at least one connected tool. If your instance uses queue mode, Simple Memory does not work in an active production workflow. Prices are listed billed annually.
- **Last checked:** Nov 2, 2026.

## 4. Lindy

Lindy is grade 2 for agents described in plain language, from Team at $49.99 a month (lindy.ai/pricing, checked Nov 2026).

- **Grade:** 2. Lindy lists "1,500+ integrations + MCP" and says "Its memory is yours to edit." Its pages do not say the agent writes that memory itself, so the grade stops at 2.
- **Plan required:** Team. Free gives $50 in credits with no credit card, and those credits last 7 days.
- **Price:** Team $49.99 a month for 5,000 credits a month. Enterprise is custom.
- **Connects to:** 1,500+ integrations. "Any hosted MCP server plugs in too."
- **Tools and MCP:** hosted MCP servers.
- **Memory:** editable memory and persistent workspace context. Whether the agent writes memory itself: Not documented.
- **Best for:** setting up an assistant from one sentence and running it on a schedule.
- **Limitation:** free credits expire after 7 days. The pricing page shows no annual price.
- **Last checked:** Nov 2, 2026.

## 5. Microsoft Copilot Studio

Copilot Studio is grade 2 for agents inside Microsoft 365, at $200 a month per pack of 25,000 Copilot Credits (Copilot Studio pricing page, checked Nov 2026).

- **Grade:** 2. Prebuilt connector tools, MCP tools and resources, and Memory, which helps an agent "remember details from its interactions and use that context for future interactions." Memory counts as Partly because it is "a production-ready preview feature" for "agents powered by the GitHub Copilot harness", and "If a user doesn't interact with the agent for 28 days, the system deletes their memories from that agent." So the grade stops at 2.
- **Plan required:** a capacity pack. Microsoft also offers a free trial.
- **Price:** $200.00 per pack per month for 25,000 Copilot Credits. Licensed Microsoft 365 Copilot users can build and use internal agents within Microsoft 365 at no additional cost.
- **Connects to:** standard connectors such as SharePoint on all plans. Premium connectors are in select plans.
- **Tools and MCP:** MCP needs generative orchestration turned on.
- **Memory:** a preview for agents on the GitHub Copilot harness. If a user does not use the agent for 28 days, their memories are deleted.
- **Best for:** Microsoft 365 organizations.
- **Limitation:** Memory is a preview for one harness only. Event triggers use only the agent maker's credentials, and enabling them can change how billing is calculated.
- **Last checked:** Nov 2, 2026.

## 6. OpenAI Agents SDK

The OpenAI Agents SDK is grade 0 as shipped (2 in code) for developers, free and open source under the MIT license (github.com/openai/openai-agents-python, checked Nov 2026).

- **Grade:** 0 (2 in code). No prebuilt app catalog or trigger is documented. In code, you get function tools and MCP. Sessions is "a persistent memory layer for maintaining working context within an agent loop". Whether it is kept across runs is Not documented, and it is code you write, so it does not move the grade.
- **Plan required:** none. Install it with `pip install openai-agents`.
- **Price:** $0 under the MIT license. No price is documented on the SDK pages.
- **Connects to:** any Python function, turned into a tool "with automatic schema generation". It supports the OpenAI Responses and Chat Completions APIs and 100+ other LLMs.
- **Tools and MCP:** remote MCP tools sit alongside function tools. Handoffs let one agent delegate to another.
- **Memory:** Sessions, "a persistent memory layer for maintaining working context within an agent loop."
- **Best for:** developers who want handoffs, guardrails and MCP in Python.
- **Limitation:** you write and host the code. Agent Builder, OpenAI's visual canvas, is scheduled to shut down on November 30, 2026.
- **Last checked:** Nov 2, 2026.

## Best for each need

- Best for no-code agents on a free plan: Make AI Agents (grade 2, Free, $0)
- Best for the widest app catalog: Zapier Agents (grade 2, Free, $0)
- Best for self-hosting: n8n (grade 2, Starter 20€ a month billed annually)
- Best for plain-language setup: Lindy (grade 2, Team $49.99 a month)
- Best for Microsoft 365 teams: Microsoft Copilot Studio (grade 2, $200 a month per pack)
- Best for developers writing code: OpenAI Agents SDK (grade 0, 2 in code, MIT license)

## FAQ

### Can I build an AI agent for free?

Yes: Make Free includes 1,000 credits a month and Zapier Agents Free includes 400 activities a month, both at grade 2. n8n's Community edition is free to self-host, and the OpenAI Agents SDK is MIT licensed.

### Do I need to code to build an AI agent?

No. Make calls itself a no-code platform, Lindy sets up from a sentence and Copilot Studio is low-code; the OpenAI Agents SDK is a Python library.

### What is the easiest way to build an AI agent?

Describe it in plain language: Lindy says "a sentence is all it takes", and Copilot Studio builds an agent from a description you test before publishing. For no code on a free plan, Make AI Agents is the pick on this page, at grade 2 on Free.

### How much does it cost to build an AI agent?

Make Free and Zapier Agents Free cost $0, while n8n Starter is 20€ a month and Zapier Pro about $33.33, both billed annually. Lindy Team is $49.99 a month and a Copilot Studio pack is $200 a month.

### Can I still use OpenAI Agent Builder?

OpenAI is deprecating Agent Builder, and it is scheduled to shut down on November 30, 2026. ChatKit remains available, and you can download the Agents SDK code to deploy a workflow yourself.

### Can an AI agent use MCP tools?

Yes: Make, n8n, Lindy and Copilot Studio document MCP tools, and the OpenAI Agents SDK exposes remote MCP tools in code. For apps that hand their own data to agents, see [MCP and API access](/best-ai-notetakers-mcp-api-ai-agents).

### What does an AI agent need to work?

A chat model and at least one tool, as n8n's AI Agent node docs state; the Agents SDK adds instructions, defining agents as "LLMs equipped with instructions and tools". Memory, such as n8n's Simple Memory node, keeps chat history for the current session.

## Sources

All pages fetched Nov 2, 2026.

- Make: [AI Agents](https://www.make.com/en/ai-agents), [pricing](https://www.make.com/en/pricing), [Introduction to AI agents](https://help.make.com/introduction-to-ai-agents), [MCP tools for AI agents](https://help.make.com/mcp-tools-for-ai-agents)
- Zapier: [Agents](https://zapier.com/agents), [pricing](https://zapier.com/pricing), [How Zapier Agents usage is measured](https://help.zapier.com/hc/en-us/articles/26559132765325-How-is-Zapier-Agents-usage-measured), [Set up your agent's trigger](https://help.zapier.com/hc/en-us/articles/45394909914381-Set-up-your-agent-s-trigger), [Migrating from Agents to AI by Zapier](https://help.zapier.com/hc/en-us/articles/47402591569805-Migrating-from-Agents-to-AI-by-Zapier)
- n8n: [pricing](https://n8n.io/pricing/), [integrations](https://n8n.io/integrations/), [AI Agent node](https://docs.n8n.io/integrations/builtin/cluster-nodes/root-nodes/n8n-nodes-langchain.agent/), [MCP Client Tool node](https://docs.n8n.io/integrations/builtin/cluster-nodes/sub-nodes/n8n-nodes-langchain.toolmcp/), [Simple Memory node](https://docs.n8n.io/integrations/builtin/cluster-nodes/sub-nodes/n8n-nodes-langchain.memorybufferwindow/), [memory overview](https://docs.n8n.io/advanced-ai/examples/understand-memory/), [hosting](https://docs.n8n.io/hosting/)
- Lindy: [pricing](https://www.lindy.ai/pricing), [home page](https://www.lindy.ai/)
- Microsoft Copilot Studio: [What is Copilot Studio](https://learn.microsoft.com/en-us/microsoft-copilot-studio/fundamentals-what-is-copilot-studio), [connectors](https://learn.microsoft.com/en-us/microsoft-copilot-studio/advanced-connectors), [triggers](https://learn.microsoft.com/en-us/microsoft-copilot-studio/authoring-triggers-about), [MCP](https://learn.microsoft.com/en-us/microsoft-copilot-studio/agent-extend-action-mcp), [Memory](https://learn.microsoft.com/en-us/microsoft-copilot-studio/agents-experience/memory-overview), [pricing](https://www.microsoft.com/en-us/microsoft-365-copilot/pricing/copilot-studio)
- OpenAI: [Agent Builder guide](https://developers.openai.com/api/docs/guides/agent-builder), [Agents SDK docs](https://openai.github.io/openai-agents-python/), [Agents SDK on GitHub](https://github.com/openai/openai-agents-python)