Make AI Agents Pricing: What One Agent Run Costs

Make AI agents run on every Make plan, the free one included, and each run costs 1 credit plus a small token charge. On Make's Medium model, a job the size of Make's own example (2,000 tokens in, 100 out) adds about 0.16 credits, so one agent run costs about 1.16 credits before any tools it calls (help.make.com, read 1 October 2026). Bolted onto a booking webhook, that is roughly 316 agent-handled bookings a month on Free and about 3,160 on Core. Start free. Move to Core when the agent becomes your third automation, when you want to plug in your own OpenAI or Claude key, or when bookings outgrow 1,000 credits.

This is a source roundup, not a test. We have not run a Make AI agent. Our own Make Free account used all its credits in September and stays at zero until 16 October, which is what our Make.com free plan test is about. The agent costs below come from Make's help centre and pricing page, each linked with the date it shows. The booking costs come from our own account. The per-run totals are our arithmetic on Make's published rates, and we label them as that.

What we read, and what we did not

What we have not done: built an agent, sent it a booking, or read a real token count from a run. So nothing here says how good the agent's replies are or how many tokens your job will really use. Make also says the app is "in open beta; product functionality and pricing may change" (17 Sep 2026). Read every number below as Make's price on 1 October 2026. How we work is on how we test.

What one agent run costs

Make bills an agent the same way it bills everything else: in credits from your monthly allowance. Its credit table for the Make AI Agent (New) app (8 Sep 2026) says:

A Make staff member put the base rule plainly in the community forum: "it's one credit per 'Run an agent' operation," and each tool the agent uses also consumes its own credits (Make Community, 9 Dec 2025). So the real cost of a run is:

1 credit for the run + token credits + 1 credit for every tool the agent calls.

The token part, in credits

Make's AI Provider offers three model sizes. Since 25 August 2026 input and output tokens are priced separately (Help: Credits, 30 Sep 2026):

Make's own example: a data-extraction step with 2,000 input tokens and 100 output tokens on Medium costs "about 0.16 credits", down from about 0.6 before the change. Here is what that means for two job sizes. These sizes are examples to show the arithmetic, not measurements from our account:

Two things stand out. The token charge is small next to the 1 credit every run costs anyway. And Large costs about five times Medium for the same job, so pick Large only when Medium's answers are not good enough. After your first real runs, Make's execution details show the actual tokens, and that number replaces our examples.

What an agent adds to a booking flow

This is where our own numbers come in. In our Make Free account, a booking that arrives by webhook and sends the client a confirmation email cost 2 credits (18 September 2026). Add an agent step that reads the booking and drafts a personal reply, and the totals look like this (our arithmetic):

Each tool the agent calls adds 1 credit per booking on top: an agent that also looks up the client in a Google Sheet and writes a calendar note costs 2 more.

Do not put an agent behind a polled trigger on Free. In our account, a Google Sheets trigger checking every 15 minutes used 998 of 1,000 credits in 10 days and 7 hours without a single booking. Checks that find nothing never reach the agent, so the agent itself is not the drain. The polling is, and it leaves nothing for the agent to spend. Use a webhook trigger, which costs nothing while it waits.

When Make Free stops being enough

Make Free gives you 1,000 credits a month and 2 active scenarios (Make calls an automation a scenario). Three walls are worth knowing before you build:

1. The third automation. Our account refused a third active scenario with "Maximum number of active scenarios has been exceeded - Payment Required" (18 Sep 2026). If you already run a booking flow and a reminder flow, an agent in its own scenario is your third. Putting the agent inside the booking scenario avoids that, at the cost of a more crowded scenario.
2. Your own AI key. Free can only use Make's AI Provider. "Users on paid plans can also use custom AI provider connections" such as OpenAI or Anthropic Claude (8 Sep 2026). With your own key, a run costs 1 credit per operation in Make and the tokens are billed by your AI provider instead. That matters if you already pay for an AI account or want a model Make's three sizes do not cover.
3. Volume. About 316 agent-handled bookings a month on Medium is plenty for one coach. A clinic or an agency running several clients' bookings through one account passes it fast, and every empty poll or extra tool call brings the wall closer.

Core costs $10.59 a month, or $9.00 a month billed yearly, for 10,000 credits and unlimited active scenarios (make.com/pricing, 18 Sep 2026). That is ten times the credits, no cap on active automations, and the option to use your own AI key. Our Make.com pricing breakdown covers what Core adds beyond agents.

How you set one up

From Make's guide Create your first AI agent (20 Jul 2026). We have not done these steps ourselves.

1. Plan the job. Decide what the agent does, which tools it needs, what it should know, and what starts it.
2. Build the scenario. Add the trigger, for example a webhook from your booking form, plus any modules you need.
3. Add the agent. Use the Make AI Agent (New) > Run an agent module. On Free you pick Make's AI Provider; on a paid plan you can pick your own OpenAI or Claude connection. Then choose the model and write the instructions.
4. Give it tools. Tools can be other modules, whole scenarios, MCP servers or sub-agents. Each call costs at least 1 credit.
5. Add knowledge. Upload reference files, such as your service list or FAQ. Uploading converts the file into tokens, which costs credits.
6. Test it. Chat with it or run the scenario, then open the execution details to see the real token use. Make's default step timeout is 300 seconds.

Who should skip Make AI agents for now

If you do not have an automation yet, start with one plain booking flow and add the agent later: our Make vs Zapier free plan test and the booking setup guide for coaches show what to build first. And if the job is a single fixed prompt with no tools and no decisions, such as "summarize this form", an agent adds steps you will not use. Keep the scenario plain.

Verdict

Make AI agents are cheap per run: 1 credit plus well under 1 credit of tokens on Medium for a job the size of Make's example. That is why the free plan is a fair place to try one. Put the agent behind a webhook, keep it on Medium, and check the real token count after the first runs. Move to Core when the agent is your third automation, when you want to use your own OpenAI or Claude key, or when your bookings pass about 300 a month. The app is still in open beta, so check Make's credit page again before you build on these numbers.