Make AI Agents is free for everyone: what it means — and what it doesn't — for small businesses

Make opened AI Agents to all plans including free, and the question isn't whether you can use them for free — it's how long before the credits run out. Since February 2026, any Make account, including the free tier, can create AI agents that choose their own tools, reason over data, and take actions across 3,000+ apps. The catch is that every reasoning step, every tool call, and every response burns credits, and the Free plan has 1,000 credits per month. For a small business processing 50 WhatsApp leads per day, that's less than a week of coverage.

At Mintec, we implement AI agents in Make and n8n for clients across Mexico, Colombia, and Central America. Make's announcement changes the entry equation but not the scaling equation. Here's what matters, what they don't tell you, and how to decide if free Make works or if you need a different path.

What actually changed with Make AI Agents free

Make didn't launch a new product. What they did was remove the access barrier: AI Agents used to require paid plans. Now, an AI agent in Make is built on the visual canvas, connected to tools (Make modules or complete scenarios), and the LLM decides at each step which tool to use. The difference from a normal scenario is that the agent reasons — it doesn't follow a fixed path but interprets the task and chooses.

What the Free plan includes:

• 1,000 credits per month
• 2 active scenarios maximum
• 15-minute minimum interval between executions
• Access to Make's AI Provider (no OpenAI or Anthropic account needed)
• AI agents with the same capabilities as paid plans

What it doesn't include:

• Custom AI provider (paid plans only)
• Priority execution
• Private Spaces (per-member spend control)
• More than 2 active scenarios

The real restriction isn't features — it's the credit budget. And that's where most small businesses underestimate consumption.

The real cost: how many credits an agent consumes

Every interaction with a Make AI agent draws credits from multiple sources simultaneously:

Real example: an agent classifying a lead from WhatsApp

1. Receives message → 1 credit
2. Queries CRM → 1 credit
3. Classifies with AI → tokens (~2-3 credits)
4. Sends Slack notification → 1 credit
5. Updates CRM → 1 credit

Total: ~6-7 credits per processed lead. With 1,000 free credits, that's roughly 140-160 leads per month. If your business gets more than 5 leads per day, the Free plan won't last the month.

Core plan real price: $10.59/mo for 10,000 credits. With the same lead-classification agent, that's ~1,400-1,600 executions per month. Enough for most small businesses with moderate volume.

The "free" trap: comparison with n8n self-hosted

Make AI Agents free is an entry point, not a scaling model. The difference becomes clear when compared with n8n self-hosted:

Mintec's opinion: Make free works for prototyping and for businesses with fewer than 5 daily interactions. But the credit model turns automation into a variable cost that scales with your success — more leads, more cost. n8n self-hosted inverts that equation: cost is fixed (the server) and executions are unlimited. For a business processing 20+ leads per day, n8n self-hosted is cheaper within 3 months.

Decision framework: Make free, Make Core, or n8n

Choose based on three questions:

1. How many agent executions do you need per month?

2. Does the agent need to access sensitive data?

If the agent handles patient, financial, or legal data, data location matters. Make stores data on US/EU servers. n8n self-hosted runs wherever you configure it — on a VPS in Mexico, Colombia, or wherever you need sensitive data compliance.

3. Do you need RAG, long-term memory, or complex multi-tool agents?

Make's Knowledge module supports documents and semantic search, but it's limited compared to n8n + LangChain + vector store. If your agent needs to remember past conversation context, query vector databases, or coordinate multiple agents, n8n gives you more control.

What Make doesn't tell you about AI credits

There's a hidden cost most guides miss: AI credits and operation credits are separate balances. The Core plan includes 2,300 AI credits per month (per ConnectSafely data, August 2026). Each LLM call draws from that balance. A conversational agent processing 100 messages daily with GPT-4o can exhaust AI credits in 3-4 days, even with operation credits to spare.

If you use Make's AI Provider (no own key): you pay both balances. Convenient, but costs double.

If you use your own OpenAI/Anthropic key: you pay only 1 Make operation credit + your API bill directly. More control, more setup, but predictable.

At Mintec, we recommend the second option for any client billing more than $5,000/mo. The cost control is worth it.

Practical case: lead classification agent for a dental clinic in CDMX

We implemented a Make agent for a dental clinic receiving WhatsApp inquiries:

• Before: receptionist manually reviews each message, classifies, responds during business hours. Average time: 8 minutes per lead. Leads lost outside hours: ~35%.
• After: AI agent classifies in real time, books appointment if lead qualifies, sends context to the dentist. Average time: 45 seconds. Leads lost outside hours: ~5%.

Real first-month cost on Make Core:

• 1,200 executions × 4 avg credits = 4,800 operation credits
• ~1,200 AI calls × tokens = ~800 AI credits
• Total: within Core plan ($10.59/mo) + included AI credits
• ROI: 35% more leads captured = ~15 additional appointments × $800 MXN average = $12,000 MXN additional revenue for $10.59 USD cost

For this case, Make Core is the right choice. Volume doesn't justify n8n self-hosted, and implementation speed (2 days vs 1 week) was the deciding factor.

When to migrate from Make to n8n

Don't migrate for the sake of it. Migrate when one of these triggers activates:

1. Credits exceed $50/mo — at that point, n8n self-hosted ($6-20/mo) is cheaper with unlimited executions
2. You need data in-region — LatAm privacy regulations or enterprise client requirements
3. Complex multi-step agents — 10+ nodes with RAG, memory, and multiple data sources
4. More than 2 active scenarios — Make Free limits to 2; Make Core unlocks them but per-credit costs still scale

The migration isn't binary. Many teams use Make for prototyping and n8n for production. Build with AI, execute with rules is a pattern that works: the Make agent tests the logic, and n8n executes it deterministically in production.

What this means for automation agencies

Make's announcement changes the sales conversation. Before, the pitch was "you need an initial investment for AI agents." Now it's "you can start free, but scaling has a cost."

For agencies like Mintec, this is an opportunity, not a threat. The client who starts with Make Free discovers automation works but hits the ceiling. That's where the agency steps in: designing the right stack for the business's real volume, not the free plan's volume.

The rule we apply: the client's stack should cost less than 1% of their monthly revenue. If the business bills $50,000/mo, automation shouldn't exceed $500/mo. Make Core ($10.59/mo) fits comfortably. n8n self-hosted ($6-20/mo) does too. What doesn't fit is a $500+/mo stack with HubSpot Enterprise for a business billing that amount.

Frequently asked questions

Can I use Make AI Agents with GPT-5? Yes, if you connect your own OpenAI key on a paid plan. Make's AI Provider includes models like GPT-4o, but for newer models like GPT-5.6 (launched July 2026), you need your own connection and pay OpenAI directly.

Does Make AI Agents replace a WhatsApp chatbot? It can work as a chatbot, but with limitations: it lacks the native WhatsApp Business API integration that n8n or a dedicated solution provides. For high-volume WhatsApp customer support, you need WhatsApp API + n8n, not just Make.

What happens if I run out of Free plan credits? Scenarios stop executing. There's no penalty, but your automation halts until next month or until you upgrade. For a business that depends on automation to capture leads, that means lost leads.