What One AI Agent Run Costs on Make’s Free Plan

Make’s free plan hands you 1,000 credits a month, and the plan comparison on its pricing page marks AI agents as included on that plan. Both of those are easy to read. The number underneath them is not: what does one agent run actually take out of the 1,000, and does the product tell you?

On October 31, 2026 we ran four measurements on the free account we have held since August 2026, compared the module panel against the run-level figures in the History tab and in Credit Usage, and re-read the pages that set the rules. This page is what the screens and the pages said.

Quick Answer

One agent run on the free plan cost 1.05 credits, and the figure matched on two readings six days apart. The same one-sentence summarising task, on the Small tier, read 1 operation – 1.05 credits – 1.5 KB on October 25 and 1 operation – 1.05 credits – 1.5 KB again on October 31. The module prints that line the moment it finishes.

A run in which the agent called one tool was billed 2.09 credits, and the panel that appears when the run ends said 1.09. Both figures are Make’s own, and both are correct — they answer different questions. The panel reports the module you are looking at. The scenario’s History tab reported the run: Credits 2.09, Operations 2. In that run, the number shown first was the smaller one.

The conversational builder runs out on a free account, and the comparison table does not say so. Make’s own Maia pricing page gives a free plan 30 days from signup or 45 messages, whichever comes first. The pricing comparison marks Maia with a check in the Free column and leaves the limit out. Our account, opened in August, was past it: Your free trial has ended.

Make’s Pricing Table Disagrees With Its Own Help Centre About Maia

Maia is the chat panel that builds a scenario for you from a description — the kind of feature a reader may have in mind when they ask whether an automation tool “has AI”, which makes its place in the free plan a purchase question rather than a detail. We looked in four places on October 31, 2026 and got two answers.

Where we read it	What it says about Maia on Free
Pricing page, plan comparison	A green check in the Free column. The row reads Maia by Make — Build automations and AI agents through natural conversation, with no asterisk and no trial wording
Help centre, introduction page	Maia is available in public beta on all paid plans. On Free plans, she’s available for a 30-day trial starting from signup.
Help centre, Maia pricing page	If you’re on a Free plan, you access Maia on a 30-day trial basis after signup, with 45 free messages included. Once you hit the limit or the trial ends, upgrade to a paid plan to continue using Maia.
The product, on our free account	Your free trial has ended. To continue building with Maia, upgrade to a paid plan anytime.

Three of the four agree that free access to Maia is temporary; the dedicated pricing page is the one that states the limit, which is 30 days or 45 messages, whichever comes first. The comparison table is the outlier, and it is the page a reader picks a plan from. The same table uses an em dash elsewhere to mark a feature as absent — Make Code App carries one in the Free column — so on that page’s own usage, the check mark in the Maia row is the mark for “you get this”. We also read the thirteen questions in the pricing page’s FAQ, which is where that page keeps its qualifications; none of them mentions Maia.

That page is also the one that says what a Maia message costs, and the figures are not agent-run figures. Maia draws on the same credit pool as your scenarios, and Make’s own worked examples describe single messages at 140.04 credits and 160.18 credits — a cost the page says is the same regardless of your plan. Those three statements do not sit together: 45 messages at the rate of those examples would be several times a free month’s 1,000 credits, and Make does not say how the free messages are accounted for. Once the trial or the 45 messages run out, its instruction is to upgrade.

What a reader should do with that: treat the comparison table as a list of features that exist, not a list of what your account will do on day 31. The agent node in the next section is a different feature from the builder, and it genuinely is on the free plan.

What We Tested

Run	Setup	What the screens said
1	Run once, Small tier, the same summarising task we used on October 25	1 operation – 1.05 credits – 1.5 KB, agent thinking 0.9s
2	Run once, Medium tier, same task	1 operation – 1.11 credits – 1.4 KB, agent thinking 1.5s
3	One message to the same agent through its Chat panel	1.05 | 0.9 sec printed under the reply
4	Run once, Small tier, with one tool attached and a prompt that forces the agent to call it	Panel: 1 operation – 1.09 credits – 1.7 KB. History: Credits 2.09, Operations 2

All four readings were taken on one free account on October 31, 2026, one reading per condition, on Make’s own AI provider.

The First Number You See Is Not the Run’s Bill

Run 4 is the one worth your attention. We attached a single built-in tool — Tools: Increment function, which needs no connection and returns a number — and told the agent to call it once and report what came back. The trace in the result panel showed exactly that: Agent was thinking 0.7s, then Tool was called: Tools: Increment function, then Agent was thinking 0.6s, then the answer, 1.

The header of that panel read 1 operation – 1.09 credits – 1.7 KB. One operation. The scenario’s History tab, one click away, recorded the same run as Manual run, Success, 1 second, Credits 2.09, Operations 2, and the organisation’s Credit Usage page logged a row of 2.09 credits at 10:04:51. The tool was the second operation, and it cost a credit of its own.

Make’s documentation supports both numbers, in two different places. The credit usage page for AI agents splits the charges into rows, and the agent and its tools are separate rows. The table has two columns, one for Make’s own AI provider and one for a custom provider connection, which is a paid-plan option; a free account is reading the first column:

Run an agent — 1 credit per operation + credits based on AI tokens
Chat — 1 credit per operation + 1 credit per operation from called tools + credits based on AI tokens
Tools — 1 credit per operation and credits based on AI tokens

Note where the phrase about called tools sits. It is printed on the Chat row, not on the Run an agent row. For a scenario run, the tool’s credit comes from the Tools row instead — the same place the pricing FAQ puts it when it lists transform data using the built-in tools among the actions that consume credits. The FAQ’s exemption list is short: error handler modules and the Router, and nothing else.

The panel is not wrong either. Make’s page on operations documents the same per-module reporting for the canvas bubble: each module’s operation count, with credits shown by a toggle next to the scenario name. In our run the result panel behaved the same way — the tool ran inside the agent module rather than as a module of its own, so the panel showed 1.09 while History showed 2.09.

This matters more on an agent than on an ordinary scenario, because calling tools is the whole point of one. Make’s table charges each tool operation on its own line, and in our one-tool run the gap between the panel and the bill was exactly one credit. The run-level figures are in the History tab and in Credit Usage, and those are the ones to read.

Both Free Tiers Are the Same Model

The model list on a free account offers two choices and locks the rest. Small is gpt-5-nano with reasoning set to minimal; Medium is gpt-5-nano with reasoning set to low. Large is gpt-5-mini and carries a Paid lock, as does every named OpenAI model in the list below it. That was the arrangement on October 25 and it had not changed on October 31.

The two free tiers are therefore the same model at two reasoning settings, and Make publishes the same conversion rate for both: 18,080 input tokens or 2,260 output tokens per credit. Our two readings of the identical task differed by 0.06 credits, 1.05 on Small against 1.11 on Medium. Make’s own description of the reasoning setting says Higher reasoning improves quality, but uses more AI tokens. Neither the result panel nor the history row prints a token count, so the gap is visible in credits and not decomposable beyond that.

Chat Costs Credits Too, and Prints the Figure Per Message

The agent canvas has a Chat panel for trying the agent before you schedule it, and on the free plan it opens and works. One message — the same text we had just run through the module — came back with 1.05 | 0.9 sec printed under the reply. That is a per-message cost, shown in place, which is more than the run panel manages for tools.

Our chat message was not filed separately. The message appeared in the scenario’s own History as a Manual run of 1.05 credits, and in Credit Usage as a row typed Scenario. The Credit Usage page keeps three buckets — Scenarios, Maia and Agents — and after four agent readings in one morning the Agents bucket still read 0 credits, as it did when we first looked at it on October 25. Whatever that bucket counts, it is not the agent runs or the agent chat we performed.

What 1,000 Credits Actually Covers

Taking our measured task at face value, a free month’s 1,000 credits is roughly 950 runs of that size, or roughly 480 if every run calls one tool. That is our arithmetic on one small task, not a figure Make publishes: a longer prompt or a longer answer moves the token component, and the tool figure rests on a single reading.

The more useful number is what the rest of the account takes first. One ordinary scheduled scenario on our free plan — a Google Sheet watched for new rows, feeding Gmail — logged 1 credit per check at the shortest interval schedule available on Free, which the pricing page gives as a 15-minute minimum interval. That is 96 credits a day, and our dashboard put the recent average at 89. A single always-on scenario, doing nothing more than checking, outruns a free month’s credits on its own, before any AI runs at all.

Running out has a defined ending. Make’s FAQ says scenarios stop until credits are added, incoming webhooks queue, and the automatic purchase of extra credits is offered on the Core, Pro and Teams plans only. Credits expire at the end of the term. For a free account that leaves two exits: wait for the monthly reset, or upgrade. At the pricing slider’s smallest setting of 10,000 credits a month, Core was listed at $9 a month billed annually and $10.59 billed monthly on October 31, 2026.

Who It’s For, and Who Should Skip It

Worth opening a free account for if you want to find out what an agent inside an automation actually does, on your own data. The node places, both free tiers run, the tools picker opens, and the runs we performed each left a timestamped row we could read back. Of the four we could press run on for AI on a free plan, Make’s was the only built-in AI that ran, and was billed, inside a scenario rather than beside one — n8n Cloud has no permanently free plan, so there was nothing to press.

Skip the free plan if what you came for is the conversational builder, because after 30 days or 45 messages, whichever comes first, it is a paid feature regardless of what the comparison table shows. Skip it too if you need Make Code, which the same table marks absent on Free and charges paid plans 2 credits per second of code execution time; if you want to bring your own OpenAI or Anthropic key, which is a paid-plan connection; or if your automation needs a scheduled check more often than every 15 minutes, which is the floor the free plan puts on interval scheduling. And budget the allowance before you build: in our setup the short agent task cost 1.05 credits, while the 15-minute polling scenario would spend the month’s credits on its own.

If you are still choosing between platforms on how the AI is metered rather than what it costs you per run, how automation tools charge for AI puts the five pricing models side by side. For the platform itself, our Make review covers the parts that have nothing to do with AI.

---

Sources

Our own runs were performed on a Make free account on October 31, 2026, with the October 25, 2026 reading quoted where we repeated it. The vendor pages below were read on October 31, 2026.

• Credit usage for AI agents — the table quoted above, which splits Run an agent, Chat and Tools into separate rows. The page prints Updated 08 Sep 2026.
• Operations — the checkmark icon (the first number) in the white bubbles above modules shows each module’s operation count, the rule the result panel followed in our run.
• Credits — credits as the billing unit, and the distinction between fixed and dynamic usage.
• Updated Make’s AI Provider token pricing model — the per-tier token rates, effective August 25, 2026.
• Make AI Agent (New) app — the settings reference, including the description of Reasoning effort.
• Introduction to Maia by Make — paid plans, with a 30-day trial from signup on Free.
• Maia by Make pricing — the two-limbed free limit (30 days or 45 messages), the shared credit pool, and the worked examples of 140.04 and 160.18 credits for a single message. The page prints Updated 15 Oct 2026.
• Schedule a scenario — The minimum length of the interval depends on your plan, and the note that Immediately exists only for some triggers.
• Extra credits — Extra credits purchasing is available to users on paid plans, which is what leaves a free account with two exits.
• n8n pricing — cited only for the five-tool comparison above: n8n Cloud has no permanently free plan, so it fell outside the free-plan comparison linked above.
• Make pricing — the plan comparison quoted above, the Free plan card, and the FAQ that carries the credit exemptions and the out-of-credits behaviour. Its figures were read in a browser, because the page refuses automated requests.
• Open a free Make account — this one is an affiliate link. The vendor pages above are not.