Best Workflow Automation Software: 10 Tools for 2026
Zapier’s Professional plan starts at $19.99 a month. Zapier Agents Pro, the AI product everyone actually wants, is $33.33 a month on top. The AI costs more than the plan it sits on.
ClickUp does the same thing. A Business seat is $12 a month. ClickUp Brain is a $9 per-user add-on, which is more than the entire $7 Unlimited seat below it. Microsoft prices Power Automate Premium at $15 per user, then sells Copilot Studio separately at $200 for 25,000 credits.
That is the shape of this whole category in 2026. The headline seat price is real, and it does not describe your bill.
What follows is ten platforms we could verify against the vendor’s own pricing page this month. We started from a list of fifteen. Five came out, and there is a section at the bottom explaining exactly why. We also cut four statistics from the previous version of this article, including one attributed to IDC that does not appear to exist. That is documented too.
The pricing trap nobody puts in the comparison table
Every platform here bills on two axes. There is a seat or plan price, which is the number in the ad. Then there is a consumption meter, which is the number that decides what you pay.
The meters are not the same unit, and that is the problem. Zapier counts tasks. Make counts credits. n8n counts workflow executions. Activepieces counts active flows. Power Automate counts Copilot Credits for the AI parts and charges per bot for the unattended parts.
A task is not a credit. An execution is not a task. n8n bills one execution whether the workflow has three steps or three hundred, which is a genuinely different economic model from Zapier billing a task per action. Any table that puts these prices in one column is comparing units that do not convert.
Second trap: several vendors here publish annual rates only. n8n, Power Automate, Activepieces and Workato all show yearly-billed figures with no month-to-month rate exposed on the page. When you see €20 for n8n Starter, that is billed annually. The monthly number is not published. We have labelled these throughout rather than dividing by twelve and pretending.
Third trap, and the sneakiest: a “starting from” price is often the lowest rung of a slider, not a plan. Zapier’s Professional at $19.99 buys the smallest task tier. The tiers run from 100 tasks a month to 2 million. Same plan name, wildly different invoice.
Best workflow automation software in 2026
1. Zapier
Still the default, still the widest app catalogue, still the one your non-technical colleague can actually use without a call.
The 2026 version leans hard on Zapier Agents, which take a plain-English instruction and run against your connected apps rather than following a fixed trigger-action chain. It works. It also has its own price tag: Agents Pro is $33.33 a month billed annually for 1,500 activities, sitting on top of whatever you already pay for the platform. There is a free Agents tier at 400 activities a month if you want to try before committing.
Verified pricing on zapier.com/pricing: Free at $0 with 100 tasks a month, Professional from $19.99, Team from $69, Enterprise on quote. A monthly and yearly toggle exists, and yearly saves 33%.
What it is bad at: cost at volume. Task-based billing punishes chatty workflows, and a multi-step Zap fires a task per action. Teams that outgrow Zapier almost always leave over the invoice, not the features. It is also the weakest option here for anything requiring real branching logic or data transformation.
2. Make
Read this bit carefully, because almost every roundup on the internet has Make wrong.
Make no longer bills in operations, and the Core, Pro and Teams plans no longer exist. The pricing page now states plainly that credits are the billing unit, and the lineup is three entries: Free, Make Plan, and Company. If you are reading a comparison that mentions Make’s “operations” or quotes a Core tier, that article is out of date.
Verified on make.com/en/pricing: Free at $0 with up to 1,000 credits a month, Make Plan at $9 a month for 5,000 credits, Company on custom pricing. Most module actions burn one credit, though AI-provider actions cost more, and the code module charges 2 credits per second of execution.
The visual builder remains the reason to pick it. You see the data moving through the scenario, branch by branch, which makes debugging a complex flow far less miserable than Zapier’s linear list.
What it is bad at: the credit repricing makes forecasting harder, not easier, because “most actions cost 1 credit” is a soft promise and AI actions are the exception. Make also has a real learning curve. The visual canvas that experienced users love is the thing beginners bounce off.
3. n8n
n8n is where the word “free” needs an asterisk, and where most articles get the licence flatly wrong.
n8n is not open source. We checked the LICENSE.md in the n8n repository directly rather than trusting a summary. The code is released under the Sustainable Use License, version 1.0, which is a fair-code or source-available licence, not an OSI-approved open source one. It permits internal business use and lets you modify the software, but it restricts commercial redistribution and hosting n8n as a service for others. There is a further wrinkle: any source file with .ee. in its filename or .ee in its directory name is excluded from the Sustainable Use License entirely and requires a paid n8n Enterprise License. So even the “self-hosted community edition” is not one uniform licence.
That matters commercially. If you self-host n8n to run your own company’s workflows, you are fine. If you plan to resell automation built on it as a hosted service, read the licence before you build a business on it.
Cloud pricing on n8n.io/pricing is quoted in euros, billed annually: Starter €20/mo, Pro €50/mo, Business €667/mo, Enterprise on quote. Those are the annual-billed figures; n8n does not clearly expose a month-to-month rate, so we are not going to invent one. Execution allowances are 2,500, 10,000 and 40,000 a month respectively. Companies under 20 employees can get 50% off Business via a startup plan.
The execution-based meter is the real story. One execution covers unlimited steps. For a workflow with forty actions, n8n bills once where Zapier would bill forty times. That single difference is why data-heavy teams migrate here.
What it is bad at: you need someone technical. Self-hosting means you own the uptime, the upgrades and the security patches. “Free” is free of licence fees, not free of labour.
4. Microsoft Power Automate
If your company runs on Microsoft 365, this is the path of least resistance, and the pricing is more layered than any other entry here.
Verified on Microsoft’s Power Automate pricing page, every figure below is paid yearly. No month-to-month rate is published.
- Power Automate Premium: $15.00 per user/month
- Power Automate Process: $150.00 per bot/month
- Power Automate Hosted Process: $215.00 per bot/month
- Copilot Studio: $200.00 for 25,000 Copilot Credits/month
- Process Mining add-on: $5,000.00 per tenant/month
Look at the gap between $15 and $200. The AI is not in the seat. Microsoft’s own page notes that Copilot Credit consumption varies by usage, which is a polite way of saying the AI bill is not knowable in advance. And the $5,000 process-mining tier is a reminder of who this product is really sold to.
What it is bad at: everything outside the Microsoft estate. Connectors to non-Microsoft apps exist but are second-class, and the licensing is complex enough that large deployments genuinely need a specialist to model the cost. If your stack is Google Workspace, this is the wrong tool.
5. ClickUp
ClickUp is a project management platform with automation attached rather than an automation platform, and for a lot of teams that is the correct trade. The workflows live where the work already is, so there is no syncing tasks between two systems.
ClickUp Brain handles the AI layer, and it demonstrates this article’s thesis better than anything else here. Seats are Unlimited $7 and Business $12 per user, billed yearly with no month-to-month rate on the page. Brain is a separate $9 per-user monthly add-on. The AI costs more than the $7 seat underneath it. Everything AI runs $28 per user a month. Both are credit-metered: Brain adds 1,500 AI credits per user per month, Everything AI adds 5,000.
What it is bad at: breadth is the weakness as well as the pitch. ClickUp does many things adequately and few things the way a dedicated tool does. Its automation cannot touch Make or n8n for complex branching, and long-running complaints about interface speed have not fully gone away. If you want to see how it stacks up in its home category, our best project management software comparison goes deeper, and top project management tools covers the wider field.
6. ActiveCampaign
An honest caveat first: ActiveCampaign is marketing automation, not general-purpose workflow automation. It belongs here only if your automation problem is mostly customer lifecycle. If you want to move data between fifteen internal systems, pick something else on this list.
Within its lane it is strong. The visual automation builder handles branching on behaviour, and the CRM is properly wired into the email side rather than bolted on.
We could not verify ActiveCampaign’s current pricing. Its pricing page renders client-side and returns no figures to our tooling, and the plans scale by contact count as well as by tier, so any single number would be misleading regardless. We are not going to print a price we cannot source. Check it on the vendor’s page for your contact volume before you commit.
What it is bad at: cost scales with your list whether or not you email it, and the feature depth means the interface has grown genuinely cluttered.
7. Workato
Enterprise integration, and the pricing page is a contact form.
We checked workato.com/pricing directly. It publishes no dollar figures at all, only “pricing designed to support organizations of all sizes” and a demo booking link. If you see a Workato monthly price in a comparison article, the writer got it from an aggregator, not from Workato.
The absence of a price is itself the useful signal. Quote-only means sales-led, which means a procurement cycle, a platform fee you will negotiate, and a floor well above what a small team spends. That is not a criticism, it is a filter. Workato is genuinely good at governed enterprise integration with proper error handling and audit trails.
What it is bad at: being affordable or quick to buy if you are under a hundred people.
8. Tray.ai
Note the name. The branding is now Tray.ai, not Tray.io; the footer on its own site reads “© 2026 Tray.ai” even though the domain still resolves at tray.io. Articles still calling it Tray.io are working from old notes.
Also quote-only. Its pricing page states that pricing is based on usage and features, and directs you to sales. The tiers are named Pro, Team and Enterprise, scoped by breadth of use rather than by feature count: single use case, departmental, then multi-department with partner integrations.
It aims squarely at revenue and marketing operations, and the fit there is real. RevOps teams that need Salesforce, Marketo and a warehouse to agree with each other get more out of Tray than out of a general-purpose tool.
What it is bad at: same as Workato. No published price, sales-led motion, overkill for small teams.
9. Pipedream
For developers who find the visual builders patronising. Pipedream lets you drop into Node, Python, Go or Bash at any step, which means you stop fighting the abstraction the moment the abstraction stops helping.
We could not verify Pipedream’s current pricing. The pricing page renders client-side and returned no plan names or figures to either our fetcher or a direct request. Rather than repeat numbers from a third-party summary, we are flagging the gap. Check pipedream.com/pricing in a browser before you plan a budget around it.
What it is bad at: it assumes you can code, and it does not pretend otherwise. Hand this to an operations manager and nothing will get built. It is also a poor fit for workflows that need non-technical colleagues to maintain them after you leave.
10. Activepieces
The entry most of these lists miss, and the one that makes the n8n licence conversation concrete.
Activepieces’ community edition is MIT licensed. Not fair-code, not source-available with restrictions. MIT. That is a genuinely open source licence, and it is the cleanest answer available if licence terms are a hard constraint for you. Where n8n’s Sustainable Use License tells you what you may not do commercially, MIT more or less does not.
Verified on activepieces.com/pricing: the cloud Standard plan is usage-based with a free entry point, then $5 per active flow per month. Ultimate is a custom annual contract. Self-hosting the community edition is free, described by the vendor as needing technical skills and covering core features only.
Per-active-flow pricing is worth sitting with. It does not care how often your flow runs. If you have a handful of very busy workflows, this is dramatically cheaper than anything task-metered. If you have two hundred rarely-used flows, it is the worst deal here.
What it is bad at: a smaller integration library than Zapier or Make, a smaller community than n8n, and the free self-hosted tier explicitly ships core features only. It is the youngest project on this list and it shows.
Comparison table
| Tool | Verified price | Meter | Best for |
|---|---|---|---|
| Zapier | Free; Professional from $19.99/mo; Team from $69/mo | Tasks (100 to 2M) | Widest app coverage, non-technical users |
| Make | Free (1,000 credits); Make Plan $9/mo (5k credits) | Credits | Visual building, complex branching |
| n8n | Starter €20, Pro €50, Business €667 (billed annually) | Executions (unlimited steps) | Self-hosting, step-heavy workflows |
| Power Automate | Premium $15/user; Process $150/bot (paid yearly) | Copilot Credits, per bot | Microsoft 365 estates |
| ClickUp | Unlimited $7, Business $12/user (yearly); Brain +$9/user | AI credits | Automation inside project management |
| ActiveCampaign | Not verifiable; scales by contact count | Contacts | Customer lifecycle automation |
| Workato | Quote only, no published price | Not disclosed | Governed enterprise integration |
| Tray.ai | Quote only, no published price | Usage, not disclosed | Revenue and marketing operations |
| Pipedream | Not verifiable from vendor page | Not verifiable | Developers who want real code |
| Activepieces | Free entry, then $5 per active flow/mo | Active flows | MIT-licensed open source |
What we cut, and why
The earlier version of this article ranked fifteen tools. Ranking fewer honestly beats ranking more badly, so five are gone.
- Kissflow – business process management, not workflow automation in the sense this article means. Different buyer, different problem.
- Pabbly Connect and Integrately – both pitched purely on being cheap. We could not verify their current pricing to the standard applied to everything else, and “cheap” is not a claim worth publishing unverified.
- IFTTT – a consumer and smart-home tool. It has been miscategorised in business automation roundups for years, including ours.
- Bardeen – the pricing page returns HTTP 200, which proves only that a server answered. We did not confirm the current product and plans to the standard we applied elsewhere, so it is out rather than in on a technicality. A live domain is not a live product.
- Camunda – a process orchestration engine for developers building applications. Genuinely good, genuinely a different category.
The statistics we removed
This section exists because the previous version of this article carried four numbers that did not survive checking. If you read it before today, here is what was wrong.
“IDC projects that by 2026, 60% of organizations will embed sustainability metrics into their digital operations.” Removed. We went looking for the publication and could not find it. IDC does publish several real “by 2026, 60%” sustainability predictions, and none of them say this. The real ones are about Chief Sustainability Officers driving AI in procurement, enterprises implementing Sustainable AI Frameworks, and large organisations requiring carbon-neutrality strategies in technology RFPs. The sentence in this article was a blend that matched no actual IDC prediction. It was also sitting in a workflow automation article for no reason anyone could explain. We have not softened it or re-attributed it to a nearby real statistic, because that would launder the same problem. It is gone, along with the “Sustainability Integration” trend section built on top of it.
“According to Gartner, by 2028, 33% of enterprise software applications will include agentic capabilities.” Removed, for a different and more boring reason. This one is very widely attributed to Gartner and may well be accurate. We could not confirm it at the source: gartner.com sits behind a challenge that blocks our verification, and every copy of the figure we could reach was an aggregator quoting another aggregator. Our rule is that we do not print research citations we cannot trace to the primary document, even when we suspect they are real.
“72% of enterprises now have AI agents deployed and operating autonomously.” Corrected rather than cut, because the underlying survey is real and it is worth being precise about. Two things were wrong. First, the survey measures enterprises using or testing AI agents, not agents “deployed and operating autonomously”. Those are very different claims, and the stronger one is not supported. Zapier’s own write-up is internally inconsistent on this point, using both phrasings on the same page, and the earlier version of this article quoted the flattering one. Second, and more importantly: this is Zapier’s own research, and Zapier is the first tool in this ranking. It was conducted by Centiment for Zapier in late October 2025, with roughly 525 responses from US executives at companies with 1,000-plus employees. Vendor-commissioned research about the category the vendor sells into is not independent evidence. Read it as a marketing artefact and it is still interesting. Read it as proof and you have been sold something.
A case study claiming “24% higher open rates and 18% increase in repeat orders”. Removed. No company, no source, no date, no way to check any of it. An anonymous case study is an anecdote wearing a suit.
What actually changed in this category
Stripped of the fabricated statistics, three shifts are visible from the pricing pages alone.
AI got unbundled and metered. Two years ago AI features were thrown in to justify a price rise. Now they are separate SKUs with their own consumption meters: Zapier Agents Pro, ClickUp Brain, Copilot Studio credits. Vendors discovered that inference costs money and stopped absorbing it. Assume the AI is an extra line item until a page tells you otherwise.
The billing unit is where competition moved. Make’s switch from operations to credits and n8n’s per-execution model are not accounting details, they are the product strategy. When feature lists converge, the meter is the differentiator.
Natural-language building is real but shallow. Every platform here will now turn a sentence into a draft workflow. In practice it produces a decent skeleton that a human then fixes. Useful. Not the same as the demo.
How to choose
Skip the feature matrix. Answer three questions.
Who maintains this in a year? If the answer is a non-technical colleague, you want Zapier or ClickUp, and you will pay for that. If it is an engineer, Pipedream or n8n costs a fraction. This is the question that actually decides it, and it is the one people skip.
What shape is your workload? Few workflows running constantly favours per-execution or per-flow pricing, which means n8n or Activepieces. Many workflows running rarely favours task-based pricing, which means Zapier. Get this backwards and you will overpay by a multiple, not a margin.
Where does the data have to live? If the answer is “on our own infrastructure”, the list collapses to n8n and Activepieces, and the licence difference between them becomes the deciding factor rather than a footnote.
On budget: under $50 a month, look at Make’s $9 plan, Activepieces, or self-hosted n8n. In the low hundreds, Zapier Professional or n8n Pro. Above that you are in quote-only territory with Workato and Tray, and the number is whatever your procurement team negotiates.
Where this sits next to other automation
Workflow automation connects cloud apps through APIs. It is not the only kind and the categories get muddled constantly.
If your problem is legacy software with no API, where a bot has to drive a user interface, you want robotic process automation instead. If it is builds, deployments and infrastructure, that is DevOps automation. Content generation inside a workflow is a job for AI writing tools, and turning calls into structured text is handled better by dedicated transcription software. Teams evaluating this alongside their project stack may also want Asana alternatives.
Frequently asked questions
Is n8n actually open source?
No. n8n uses the Sustainable Use License v1.0, a fair-code or source-available licence, not an OSI-approved open source one. You can read the code, modify it, and run it for your own business. You cannot freely resell it or host it as a service for others. Files marked .ee. are excluded from that licence entirely and need a paid Enterprise License. If you need a genuinely open source option, Activepieces’ community edition is MIT licensed.
Which is cheaper, Zapier or Make?
Make’s entry plan is $9 a month for 5,000 credits against Zapier’s Professional from $19.99. But they meter different things, so the sticker comparison is close to meaningless. Zapier bills a task per action, Make bills credits per module action with AI actions costing more. Model your actual workflow volume against both. For multi-step workflows Make is usually cheaper; for a small number of simple two-step automations the gap narrows.
Is there a genuinely free workflow automation tool?
Yes, with conditions. Make’s free tier gives 1,000 credits a month. Zapier’s gives 100 tasks. Activepieces has a free entry point and a free self-hosted community edition. n8n’s community edition is free to run yourself. The free cloud tiers are real but small, and the self-hosted ones trade licence fees for your time on hosting and maintenance.
Why can’t I find pricing for Workato or Tray?
Because they do not publish it. Both pricing pages are contact forms with no dollar figures. That is a deliberate sales-led model aimed at enterprise procurement. Any article quoting you a specific Workato or Tray monthly price took it from an aggregator, not the vendor.
Does the AI cost extra?
Usually, yes, and often more than the plan it attaches to. Zapier Agents Pro is $33.33 a month on top of a platform plan starting at $19.99. ClickUp Brain is $9 per user on top of seats starting at $7. Microsoft sells Copilot Studio at $200 for 25,000 credits separately from the $15 Power Automate Premium seat. Budget for AI as its own line item.
What is the difference between a task, a credit and an execution?
A task (Zapier) is one action inside a workflow, so a ten-step Zap burns ten. A credit (Make) is roughly one module action, though AI actions cost more. An execution (n8n) is one complete run of a workflow regardless of how many steps it contains. This is why a forty-step workflow can cost forty units on Zapier and one on n8n.
Is Power Automate included with Microsoft 365?
Limited capability comes with some Microsoft 365 licences, but the standalone plans exist for a reason. Power Automate Premium is $15 per user a month paid yearly, and unattended bots are priced separately at $150 or $215 per bot a month. Copilot Studio and process mining are further charges again.
Are those monthly prices or annual?
Read carefully, because vendors default to displaying annual rates. n8n’s euro figures, ClickUp’s seats, Power Automate’s whole table and Activepieces’ Ultimate tier are all annual-billed. Zapier and Make expose a real monthly and yearly toggle. Where no month-to-month rate is published, we have said so rather than dividing by twelve.
Can these tools replace an RPA platform?
Not usually. Workflow automation talks to applications through APIs. RPA drives a user interface the way a person would, which is what you need for legacy software with no API. Power Automate is the one entry here that does both, via its per-bot Process plans. Otherwise they solve different problems.
The verdict
Most teams should start with Make’s $9 plan or Zapier’s free tier and find out what they actually need before paying for anything. Technical teams with real volume should go straight to n8n and accept the hosting work, reading the licence first if they intend to resell. If licence terms matter, Activepieces is MIT and n8n is not.
Microsoft shops should use Power Automate and budget the AI separately. Everyone considering Workato or Tray already knows who they are, and already has a procurement process.
One habit is worth more than this entire ranking. Before you buy, open the vendor’s pricing page yourself, find the monthly toggle, and work out what the meter charges for the workflow you actually intend to run. Four numbers in the previous version of this article did not survive that treatment. Neither will most of what you read about this category.