Make Review (2026): Is It the Right Automation Tool for Your Business?
Make is a visual automation platform with 3,000+ apps that recently added a native AI agent framework. It suits teams that value a fast start and a clear visual builder and are comfortable counting cost in credits rather than executions. On the free plan, which has no time limit, you can test an AI agent without a separate API key, because Make provides a built-in AI provider. The limitation: in our test on a real account, the Code tool for an agent required a paid plan, while n8n has no such restriction.
If you haven’t decided yet whether your business needs automation at all, start with our guide “What Is Business Automation and Where to Start.”
Who Make Is For
- Teams that want to reach a first working scenario fast. In our test, automatic data-structure detection produced fields ready to map within seconds.
- Anyone who wants to try an AI agent with no setup. The built-in “Make’s AI Provider” works immediately, even on the free plan, with no need for your own OpenAI or Claude key.
- Agencies with a moderate volume of client automations that find it easier to count in credits than to track executions or operations separately in every tool.
- Teams without a developer. The core workflow builder works without code.
Who Make Isn’t For
- Custom code in an AI agent without paying extra. Running a scenario with a Code tool attached to an agent fails with the error “The scenario cannot run because it contains an app for paid plans only”, which blocks the run. It is not a warning but a full block.
- Hosting the platform on your own server. Make is a cloud service and has no fully self-hosted version, unlike n8n. There are partial options, but they do not replace self-hosting. The on-premise agent (Enterprise only) is a local connector to systems in your internal network, not a deployment of the platform itself. You can choose a data-processing region between the EU and the US (according to a Make staff member in the community, data in the EU zone is stored on EU servers). If you need full control over infrastructure, Make will not fit. If a region choice is enough, check the terms in Make’s documentation: we did not verify this in our test.
- Teams that need predictable costs regardless of scenario complexity. In the credit model, the price of one run depends on the number of modules and the amount of data processed, so it is not fixed in advance. A complex multi-module scenario can use a noticeable share of the monthly credit allowance.
- Teams that need a fully autonomous AI agent. At least for the Code tool, input values are set at design time instead of being decided by the model at call time (more on this below).
What We Tested
We tested Make on a real account and completed two tasks:
- Qualifying an inbound lead through a classic scenario. A custom webhook receives signup-form data, a Router splits the flow into two branches by budget and company size, the hot branch calls Anthropic Claude (Simple Text Prompt) to draft a personalized email, and the cold branch logs a status with the built-in Set variable tool.
- The same scenario through the native Make AI Agent (a separate, newer feature in the “AI Agents” section, in open beta at the time of testing). The agent receives the lead data directly, classifies the lead and writes the email itself. We first attached a custom Code tool for scoring, then removed it once it became clear that the Code tool for agents requires a paid plan.


Core Features
- A visual scenario builder with modules.
- 3,000+ apps, available across the Make ecosystem. Individual features and limits depend on the plan.
- Make AI Agents (in open beta at the time of testing). A dedicated section of the platform: an agent with Make’s own AI provider (no external key needed) or your own key (BYOK), with pluggable Tools, Knowledge (knowledge bases) and MCP connections.
- Maia helps build automations and agents through a conversation with AI.
- Automatic webhook data-structure detection. After one test request, every field is ready to map through clickable tokens.
- Router with filters on each branch and a “fallback route” option (the equivalent of “else”).
- Credit-based pricing. For standard module actions, each action costs from 1 credit. AI operations can be metered differently depending on how much work the model actually does. The Free plan includes 1,000 credits a month.
What Works Well
Data-structure detection. We sent a test JSON payload to the webhook, and within seconds Make showed “6 values detected and ready to map”. Every field was ready for click-to-insert mapping anywhere in the scenario, with no extra setup.
The built-in AI provider. For the first agent run we needed no external account or API key: we picked “Make’s AI Provider (default)” and the agent worked immediately. The interface warned clearly: “Some models are available only on Make paid plans. If you’re on the free plan or a trial, your scenario will run using a free model.” That is exactly what happened, and it worked well. For anyone just testing an idea, it removes a separate step of signing up and getting keys.
An AI agent with no tools attached. The agent checked on its own whether the lead met the Hot-category thresholds (budget, headcount), explained its logic in three points, and wrote an email with a subject line, a greeting by name and a concrete call proposal. In our view, that text needs only light editing.
Clickable field tokens. In n8n, an expression that looked identical behaved differently depending on how it was entered. In Make, every token we inserted with a click behaved the same in all the fields we tested, from a Router filter to a tool’s input parameters.
The fallback branch in Router. When we found that both branches were firing at once (see “What Doesn’t Work”), the fix took one click: mark the second branch as a “fallback route”. After that, the branches became mutually exclusive, as we had expected from “else” logic.
What Doesn’t Work
Router branches are not mutually exclusive by default, and that is not obvious at first. We built a Router with a “Hot Lead” filter on the first branch and left the second branch with no condition, expecting the natural “else” logic. Instead, both branches ran for the same lead. A branch without a filter always runs, regardless of what other branches do, unless you explicitly mark it as a “fallback route”. It is easy to fix once you know, but the mistake stays invisible until you check the execution log by hand.
In AI agent scenarios, the webhook only starts listening after you save the whole scenario. In a classic scenario, a test request to the webhook is picked up right after the module is created. In an agentic scenario, an identical request returned 410 “There is no scenario listening for this webhook”. We had to save the entire scenario (to get a real ID) before a repeat request worked. It is a difference between the two scenario types that is easy to miss.
Code as a tool for an AI agent is fully blocked on the Free plan. This is not a “consider upgrading” hint. Trying to run the scenario produces a hard execution error: “The scenario cannot run because it contains an app for paid plans only. To use the app, please upgrade your plan.” In n8n, the Code tool for an agent is not tied to a paid plan, while in Make, custom logic inside an agent’s tool is already a reason to upgrade.
For the Code tool, input values are set statically instead of being decided by the agent. The “Value” field for a Code tool parameter is labeled “Value to pass through”: the value is mapped from scenario data at design time. The agent decides whether to call the tool, but it does not build the call arguments itself, as it did in n8n. Standard module tools in Make have a “Let AI Agent decide” toggle that lets the agent fill in fields automatically, so we confirmed this limitation only for the Code tool we tested. For simple scenarios the difference is barely noticeable, but it is worth considering for more complex agent logic.
What Make Costs
| Plan | Price | Credits/mo | What’s Included |
|---|---|---|---|
| Free | $0/mo | 1,000 | 2 active scenarios, visual builder, 3,000+ apps, minimum 15-min scheduling interval |
| Core | $9/mo (for 10k credits) | 10,000 | Unlimited active scenarios, 1-min scheduling interval, Make API access |
| Pro | $16/mo (for 10k credits) | 10,000 | Priority execution, custom variables, full-text log search |
| Teams | $29/mo (for 10k credits) | 10,000 | Team roles, shared scenario templates |
| Enterprise | Custom | Custom | Custom functions, enterprise apps, 24/7 support |
Checked: October 26, 2026. Source: the official make.com/en/pricing page. Prices are for annual billing (Make shows a 15% discount for annual payment, so monthly billing costs roughly 18% more).
Is it worth it? The Free plan has no time limit, while n8n.cloud offers only a 14-day trial. You can test an AI agent on it at no cost. But the cost in Make depends not only on the plan but also on how complex your scenario is. In our test, one AI agent call that generated an email cost 1.86 credits. At that rate, the Free plan’s 1,000 credits cover roughly 530 such calls a month, or about 18 a day. The other modules in a scenario will use additional credits, so for real use you need to count the whole scenario, not just the AI call.
Real-World Use Cases
Qualifying inbound leads for an agency. A signup form, a webhook, branching by budget and company size, then a personalized AI email for hot leads. This is the exact scenario we rebuilt and tested, including fixing a real mistake in the branching logic.
Testing an AI agent idea without a separate API key. The built-in AI provider lets you test an agent without connecting an external AI vendor. It is convenient at the early check stage, when there is no need yet to set up paid API keys or more complex infrastructure.
Simple automation for teams without a developer. The core builder works without code, and automatic webhook-structure detection makes the first steps easier, so Make fits handing simple, repeatable processes to people without a technical background.
We built the CRM, customer support and lead scoring scenarios on n8n, not on Make: see Automation Stack for Freelancers, AI Customer Support Agent Stack and AI Lead Scoring Stack for B2B.
Integrations
In our test we used Webhooks, Anthropic Claude (Simple Text Prompt) and the built-in Make AI Provider. For the R StackAI CRM cluster, HubSpot and Pipedrive matter: according to Make’s official pages, both have native apps maintained by Make, but we did not test them here. For services without a ready-made module, there are the generic Webhooks and HTTP modules. In total, more than 3,000 apps are available in the Make ecosystem. Individual features and limits depend on the plan. For Claude as a standalone tool, its pricing and limits, see the Claude Review.
Alternatives to Make
- n8n has a self-hosted option (the free Community Edition) and, in our test, did not tie the Code tool for agents to a paid plan. But with self-hosting you set up and maintain the server yourself, and the cloud version has only a 14-day trial. Detailed comparison: n8n vs Make: which one should an agency that resells automation to clients choose. To see n8n on a real scenario, read the tutorial How to Build Your First Automation in n8n.
- Zapier was tested on the same scenario in its own review: Zapier Review. Usage there is counted in tasks: each Zap step that completes an action uses a task, and the free plan is limited to 100 tasks a month and two-step Zaps. So cost grows with the number of actions.
FAQ
What is Make (make.com)?
Make is a visual no-code/low-code platform for building and automating multi-step workflows that connect apps and APIs. It has 3,000+ apps and a native AI agent framework, and its cost is counted in credits rather than executions.
Is Make really free, or is it a disguised trial?
Really free, with no time limit: 1,000 credits a month and 2 active scenarios. During testing we saw no banners about a trial ending and no day counters.
Can you build an AI agent on the Free plan?
Yes. In our test, a basic agent worked without an external API key and returned a well-reasoned response. The limitation applies specifically to the custom Code tool for agents, which requires a paid plan.
How many credits does a typical scenario use?
It depends on the number of modules and the amount of data processed. In our test, one AI agent call cost 1.86 credits, and a full scenario with several modules will cost more. It is better to estimate the expected volume up front than to find out after the limit runs out.
Does the Router split leads between branches with “if/else” logic automatically?
Not by default. A branch with no filter of its own always runs, in parallel with any other branch whose condition matches. To get real “else” logic, you need to explicitly mark that branch as a fallback route.
Do you need a developer to work with Make?
For most scenarios, no: the workflow builder works without code. You need one only if you plan to write custom code through the Make Code module, and even then that capability for AI agents is not available on the Free plan.
Is there a self-hosted version of Make?
There is no fully self-hosted version: Make is a cloud service, unlike n8n. There is an on-premise agent for connecting to systems in your internal network, but it is available only on the Enterprise plan and does not move the platform itself onto your server. You can choose a data region between the EU and the US.
Final Verdict
Choose Make if a fast start, a visual builder and the ability to test an AI agent with no separate API key and no cost matter to you. It is a convenient option for teams without a developer and with a moderate volume of automations. The price is a credit model in which cost depends on scenario complexity, and some AI features tied to a paid plan: in our test, the Code tool for an agent did not run on Free. So if you need custom logic in an agent's tools on no budget, or hosting on your own server, look at n8n instead.
- Executive Summary
- Make is a visual automation platform with 3,000+ integrations and a built-in free AI provider; best suited to teams ready to accept credit-based pricing instead of paying per execution.
- Best For
- Teams that want a fast visual start and a built-in AI agent with no external API keys, and are comfortable paying in credits rather than per execution.
- Avoid If
- Companies that need a self-hosted option, free custom code for an AI agent, or fully predictable costs regardless of scenario complexity.
- Pricing Summary
- Permanently free (1,000 credits/mo) up to $29/mo (Teams, annual billing); Enterprise is custom-priced - cost tracks credits, not a fixed tier.
- Version / Plan Tested
- Make, real account, September 2026