Disclosure first: we make BizOps Pilot, which competes with Make. Read everything below with that in mind. What we can offer in exchange is that every fact here is sourced to Make's own announcement, Make's own pricing page, or Make's own help centre, with the date we read it. Where the evidence is a user's claim rather than a vendor's statement, it says so in the sentence, not in a footnote. Where Make is better than us, it says that too.

This is not a piece about Make being greedy. It is a piece about a structural thing that happened across this entire category in the last year, which almost nobody has written down plainly, and which Make happens to have documented better than anyone else did.

What actually happened

On 13 August 2025, Make posted an announcement in its own community News section. On 27 August 2025, the change went live. Make replaced operations with credits as its billing unit.

Here is the announcement in Make's own words, from Introducing credits: a new system of billing:

"We're transitioning from operations to credits as our new billing unit."

"This update allows us to introduce different credit values for different actions, meaning your billing will better reflect the specific actions you're performing."

"The conversion is automatic... Your current plan and price stay exactly the same."

"Most actions continue to consume 1 credit per module run, where advanced features, especially AI, will be priced at a fixed rate or varying by model and token usage."

"Your credit usage may increase if you use Make's AI Provider."

"Make AI Agents: Available on all plans with Make's AI Provider. Building with your own connection will require the Pro plan or higher."

Read those six lines together and the whole thing is right there, stated by the vendor, in public, before it shipped.

The price did not move. The unit moved. And the unit moved specifically so that it could stop being one thing. Under operations, a module run was a module run. Under credits, "different credit values for different actions" is the explicitly stated purpose of the change, and Make named the actions it had in mind: AI. Make's help centre confirms the terminology is settled, on its Credits page, updated 19 January 2026.

We want to be exact about what Make's own concession says, because it is the load-bearing sentence and it is easy to overstate. Make did not say your bill will go up. Make said "your credit usage may increase if you use Make's AI Provider." That is a conditional about a specific product surface, not a confession. It is also, read carefully, an acknowledgement that the same work can now cost a different number of credits than it used to, depending on what kind of work it is.

The pattern: nobody raised a price in 2026

Here is the part that made us want to write this at all.

We went looking for 2026 price increases across five automation vendors: n8n, Zapier, Make, Gumloop and Lindy. Every number below is dated and sourced to a vendor's own page.

We found no instance of a headline plan price going up. Not one, across the five vendors we checked. Every pricing change we could source did one of three things: it repriced the meter, it added a multiplier, or it moved the unit.

Make moved the unit on 27 August 2025.

Zapier added the multipliers. On 15 June 2026, per Zapier's own help centre, AI by Zapier steps became priced by model tier: Standard 1x, Advanced 3x, Premium 5x, with new steps defaulting to Advanced, which is the 3x tier. The same page offers "connect your own AI account (1x)" as an alternative. Eleven days earlier, on 4 June 2026, Zapier documented that each successful MCP tool call consumes 2 tasks, at a fixed rate, drawn from the same plan allowance as your Zaps.

Zapier's plan prices did not go up either. Free is still 100 tasks. Professional still starts at $19.99. What changed is how much of your allowance a given piece of work consumes.

So: the price increases in this category in 2026 were not price increases. They were unit redefinitions. That framing is more accurate than "Make put prices up," which is false and which we are not going to say. It also happens to be more damaging, because you cannot see it on a pricing page.

The fair part, which we are not going to skip

Changing a unit definition is a legitimate business decision, and vendors are entitled to make it.

Here is the honest case for what Make did. When AI arrived, one module run stopped being one comparable thing. A module that writes a row into a Google Sheet and a module that runs a large language model over a long context are not the same amount of work, they are not the same amount of cost to Make, and pricing them identically was the thing that was actually wrong. Make's stated reasoning, that billing should "better reflect the specific actions you're performing," is not spin. It is a real argument, and on the merits it is a decent one.

Make also announced it publicly, in advance, on their own community, two weeks before it shipped, and said out loud that AI usage may consume more. That is more disclosure than a lot of this industry manages. We are not aware of anything hidden here in the sense of concealed. The information was published.

So the takeaway we want you to leave with is structural, not moral. When the unit is defined by the vendor, the price of your work is whatever the vendor later decides the unit means. That is true of every metered vendor, including the ones behaving perfectly well. It is a property of the pricing model, not a character flaw in the company.

A flat price has no unit to redefine. That is the entire argument, and it is worth exactly as much as it is worth to you and no more.

What one credit is, and why the number is hard to hold on to

Make's pricing page defines it:

"Each module action in your scenario, like adding a Google Sheet row or fetching Gmail account data, counts as one credit."

That is per module action, not per scenario run. So a scenario with twenty modules, run once, is in the region of twenty credits. Loops and iterators multiply from there. This is the granular end of the metering spectrum. For contrast, n8n counts one execution per whole workflow run regardless of step count, which is a materially more generous unit, and we will say so plainly because it is true.

At the entry tier of Make's pricing slider, read on 16 July 2026 and unchanged when we checked again on 22 July 2026: Free is $0 with 1,000 credits a month, and Core includes 10,000 credits a month. Make's free plan has no time limit, which is genuinely unusual and genuinely good. It does throttle you to a 15 minute minimum interval between runs.

Note what "different credit values for different actions" does to the number above. Ten thousand credits a month means ten thousand module actions only if every action costs one credit. Make's own announcement says advanced features, especially AI, will vary "by model and token usage." So the allowance is denominated in a unit that is no longer constant across the work you might do with it. That is not a trick. It is just genuinely harder to forecast than "10,000 things."

The 19,000 credit run, labelled honestly

There is a number in this story that gets passed around, and we are going to handle it carefully, because we nearly got it wrong ourselves.

On 3 June 2026, a user posted in Make's community forum in a thread called Need help with agentic workflow that a single agent run, using Make's AI provider, consumed 19,000 credits. They called the credit usage "absolutely ridiculous." Core includes 10,000 credits a month. If accurate, that one run is roughly two months of allowance.

This is a single user's report on a forum, not a statement by Make and not a number we measured. It is on Make's own community site, which is why it is worth reading, but the claim belongs to the user, not to the vendor. We are citing it as a user report and you should weigh it as one.

And there is a counter-argument in the same thread that deserves to travel with the number. A replying community member attributed the burn to the non-deterministic agent wasting tokens working out where to go, plus context bloat. That is a fair technical critique, and it matters: it means 19,000 is a worst case from a badly behaved agent loop, not what a typical Make scenario costs. Anyone quoting this number as "what Make costs" is misleading you, and we would rather say that than borrow the number's shock value.

What the report does establish, and this part is solid, is the shape of the risk: under a per-action credit meter, an agentic AI workload has no natural ceiling, because the agent decides how many actions to take. You cannot forecast a bill whose input is a model's judgement about how much work to do. Zapier evidently agrees, because their fix for the same problem is a circuit breaker: per their 15 June 2026 help page, a step that reaches 75 tasks in a single run pauses the Zap and asks you to approve before continuing. That is a good feature. It is also an admission of what AI steps do to a meter.

Bring your own connection, gated to a paid plan

One more line from the 13 August 2025 announcement, because it belongs to the same pattern:

"Building with your own connection will require the Pro plan or higher."

That was the announcement's own wording. Make's current help centre describes the same restriction more broadly, as gated to paid plans rather than naming the Pro tier specifically. Either way, the shape of it has not moved: Make's AI provider is available on all plans, and your own connection is not available on the free one.

We want to be measured about this too, because bringing your own key is not the differentiator people think it is, and we would be embarrassed to pretend otherwise. Every serious vendor in this category now offers some version of it:

  • Zapier prices "connect your own AI account" at 1x against a 3x default tier. It is a documented, first-class discount at the market leader.
  • Gumloop gives you 50% off AI model credits for supplying your own key. You pay the model provider directly and still burn half the credits. Pro plan or higher.
  • Make allows your own connection, gated to paid plans.
  • n8n self-hosted is the real thing: bring your own key, and their docs state no credit limits apply.

So "bring your own key" is table stakes. Three of those four hand you the key back and keep the meter running. The question is not whether a vendor lets you supply your own key. It is what they count after you do.

Where Make is genuinely better than us

This is the section that makes the rest of the piece worth reading, so we are not going to be grudging about it.

Make's visual builder is the best in this category, and we are not close. Routers, iterators, aggregators, error handler branches, and deep data mapping, all manipulable directly on the canvas. If you think visually and your scenarios have real branching complexity, Make will let you express things that would be awkward in our Systems canvas. That is a mature product built by a large team over many years, and there is no version of us that catches it soon.

Make's app catalogue is far larger than ours. We are not putting a number on theirs here, because we have not verified one from their own page to the standard the rest of this piece is held to, and we would rather leave a gap than fill it with a figure we half remember. The direction is not in doubt. If your stack has something regional, vertical or obscure in it, go and check Make's catalogue and check ours, and let that decide it. If they have your tool and we do not, buy Make. That is not a concession, it is just the right answer.

Make's free tier is genuinely strong. $0, 1,000 credits a month, and no time limit on it. A lot of "free" tiers in software are trials wearing a disguise. That one is not.

What we do instead, and what it costs us

Short, because you can smell a pitch.

We charge $39 a month, flat, for Pro. There is a Founding option at $249 once, hard capped at 300 seats and enforced at checkout. There is a free beta tier.

We do not meter runs, on any plan, including the free one. There is no unit to redefine, because we do not count anything. The free tier bounds how many Pilots and Systems run at once, which is three and one respectively. That is a concurrency ceiling, not a volume meter. An active Pilot on the free plan can run forever, as often as its trigger fires, and nothing counts down. The REST API and the MCP server are a Pro feature.

AI inference runs on your own key at zero markup. You pay Anthropic or OpenAI directly, at their price, with nothing added by us. This is not generosity, it is the thing that makes a flat price survivable at all: the volatile cost in an AI workflow is the tokens, and we are not in the business of buying tokens and reselling them to you inside a unit we define.

The honest costs of that, stated plainly:

  • You have to go and get an API key. That is a real step and a real annoyance, and Make's AI provider does not ask it of you. If you want one bill and never to think about a provider account, Make's model is more convenient than ours and the credit multiplier is what you are paying for that convenience.
  • n8n self-hosted beats us, and it is not close. Bring your own key, no metering, no limits, for the price of a small VPS, which is less than $39. If you are comfortable running a container and keeping it patched, that is a better deal than we are and we are not going to pretend otherwise. We win on no server to run, no upgrades to apply, no maintenance, MCP-native out of the box, and a founder who answers his email. That is a hosting and time argument, not a pricing one.
  • We are pre-launch and small. We have no customer counts to show you, no revenue figures, and no testimonials, because we do not have them yet. Make is a real company with a real support organisation. A flat $39 from a small team is a reasonable thing to be suspicious of, and "no surprise bill, low switching stakes" only partly answers it.

What we could not verify

In the spirit of the rest of it.

Make's monthly versus annual toggle state, now resolved. On 16 July 2026 we could not tell which toggle state had produced the figures we read. On 22 July 2026 we went back and clicked the toggle directly. At the 10,000 credit tier, monthly billing is Core $10.59, Pro $18.82, Teams $34.12. Annual billing, shown per month, is Core $9, Pro $16, Teams $29, a consistent 15 percent saving at every tier. The credit counts, 1,000 free and 10,000 at Core's entry tier, are what this piece actually leans on, and those were unchanged across both reads.

A third party claims Core is $10.59 a month billed annually with monthly billing "roughly 30% more," and that extra credit packs run about $11 per 10,000 with a 25% markup since late 2025. Now that the toggle is resolved we can say precisely where the first half of that breaks: $10.59 is Make's monthly price, not the annual one, and the annual saving is 15 percent, not 30. We are marking the credit pack figure unverified and not relying on it, and recording them here only so that nobody re-finds that article and trusts it.

Make's app catalogue count. Not verified against Make's own page to this piece's standard. Omitted rather than guessed.

The 19,000 credit figure is a user's forum claim, as flagged above at length. Not a Make statement, not a measurement of ours, and characterised in its own thread as an agent behaving badly.

The one thing to take away

If you are choosing an automation tool, the headline price tells you almost nothing. What you need to know is:

  1. What is the unit? Not the price. The unit. A task, a credit, an operation and an execution are not comparable quantities, which is why most price comparisons in this category are nonsense.
  2. Who defines the unit, and can they redefine it? Make redefined theirs on 27 August 2025 and told everyone in advance. That is the good version of this. The bad version is a vendor who does it quietly.
  3. What happens to the unit when AI enters the workflow? Make's answer, in their own words, is that your credit usage may increase. Zapier's answer is 3x by default. Both answers are published. Both are honest. Both mean the same work now costs more of your allowance than it did.

Then price your own three highest-volume workflows against each unit. Count the modules. Estimate the runs. The answer usually falls out immediately, and it is often not the tool with the lowest number on the pricing page.

Sources

Make's pricing was first observed 16 July 2026 and reconciled against a second, live-interacted read on 22 July 2026, which is where the exact monthly and annual dollar figures and the paid-plan connection gate below come from. Primary sources are the vendor's own announcement, pricing page, or help centre.

  • Make, operations to credits announcement, all quoted lines, announced 13 August 2025, live 27 August 2025: https://community.make.com/t/introducing-credits-a-new-system-of-billing/89480
  • Make, credits terminology, page updated 19 January 2026, read 22 July 2026: https://help.make.com/credits
  • Make pricing, credit definition, plan prices and free tier, read 16 July 2026 and reconfirmed with the monthly/annual toggle resolved 22 July 2026: https://www.make.com/en/pricing
  • Make, credit usage for AI Agents, confirms custom AI provider connections are gated to paid plans, read 22 July 2026: https://help.make.com/credit-usage-for-ai-agents
  • Make community, single user report of 19,000 credits in one agent run, 3 June 2026, plus the community's counter-analysis: https://community.make.com/t/need-help-with-agentic-workflow/110223 (user claim, not a vendor statement)
  • Zapier, AI by Zapier model tier pricing, Standard 1x / Advanced 3x default / Premium 5x / own AI account 1x, and the 75 task per step pause, 15 June 2026, reconfirmed unchanged 22 July 2026: https://help.zapier.com/hc/en-us/articles/46597632373389-AI-by-Zapier-new-model-based-pricing-starting-June-15-2026
  • Zapier, MCP usage, 2 tasks per successful tool call, 4 June 2026, reconfirmed unchanged 22 July 2026: https://help.zapier.com/hc/en-us/articles/45645738385805-How-Zapier-MCP-usage-works
  • Zapier pricing, Free 100 tasks, Professional from $19.99, read 16 July 2026, reconfirmed unchanged 22 July 2026: https://zapier.com/pricing
  • Gumloop, BYOK gives 50% off AI model credits, Pro plan required, undated vendor doc, reconfirmed unchanged 22 July 2026: https://docs.gumloop.com/core-concepts/ai_models
  • n8n, self-hosted brings your own API key with no credit limits, undated docs, and the execution definition, reconfirmed unchanged 22 July 2026: https://n8n.io/pricing/

Pricing in this category changes without notice, and the entire point of this piece is that it can change without the pricing page changing. Every number above is a snapshot with a date on it, not a promise. Re-check before you buy anything, including ours.