--- title: "Best Free AI Workflow Automation Tools in 2026, Compared by What the Free Plan Includes" description: "Nine platforms have a permanent free plan in 2026. What separates them is not price but what the free number counts: runs, steps, modules, or credits." url: https://flowrunner.ai/blog/best-free-ai-workflow-automation-tools date_published: 2026-09-15 date_modified: 2026-09-15T12:12:16-07:00 --- Comparisons September 15, 2026 11 min read # Best Free AI Workflow Automation Tools in 2026, Compared by What the Free Plan Includes Nine platforms have a permanent free plan in 2026. What separates them is not price but what the free number counts: runs, steps, modules, or credits. ![A FlowRunner Free Plans Compared card on a plum field: the headline Nine free plans. One question., the subhead What does the free number count?, and a grid pairing each meter with the tools that count that way: whole run for FlowRunner, Activepieces, Stack AI, and n8n; action step for Zapier; module run for Make; compute credit plus AI tokens for Pipedream; AI message for Dify; live workflows for Tines. Caption: No AI meter on FlowRunner Free. Your own keys.](https://flowrunner.ai/images/blog/best-free-ai-workflow-automation-tools-hero.webp) The question is never which automation tool is free. Nine of them are. The question is what the free number counts, because that is the fact that decides whether a free plan is a working platform or a demo with a meter attached. Nine automation platforms offer a permanent free plan in September 2026: FlowRunner, Activepieces, Stack AI, Zapier, Make, Pipedream, Dify, Tines, and n8n if you self-host. They differ less in price, which is zero on all nine, than in what the free number counts: a whole run, an action step, a module, a compute credit, or an AI message. They differ again in whether AI agents and human review are included in the plan or metered as a second currency. Every number below was verified on September 15, 2026 from the vendor’s own pricing page. Where a vendor does not publish a fact, this page says that instead of guessing. Here is the part a list like this one will not say out loud: the ranking is the least useful thing on the page. Every entry costs $0, so a ranking is a writer’s taste dressed as a verdict. The comparable fact is the unit. Two plans that both advertise 100 are not the same plan when one counts whole runs and the other counts action steps, and no number of stars beside the name repairs that. ## How to read a free plan Five different things get counted across this field. Learn the five and the whole category stops being confusing. - **A run.** One complete workflow execution counts as one, however many steps it contains. FlowRunner, Activepieces, Stack AI, and self-hosted n8n count this way. - **A step.** Each successful action inside a run counts on its own. Zapier’s task is this unit: a five-step Zap uses five tasks per run. - **A module over a bundle.** Make counts one credit per module run per bundle of data, which is why Make documents a three-module scenario processing ten items as roughly 31 credits. - **Compute time.** Pipedream’s credit is 30 seconds of compute at 256 MB. The meter moves with how long the code runs rather than how many steps are drawn on the canvas. - **AI itself.** Dify counts message credits. Tines grants a one-time $50 AI allowance. Zapier Agents counts activities on a meter separate from Zapier tasks. Two more questions belong beside the unit, and both get asked too late. The first is whether AI is a second currency. On most of the AI-native field it is: the plan sells credits that bundle the model cost, so the bill moves when the model does. FlowRunner uses [BYOK](https://flowrunner.ai/concepts/byok), a model where the customer supplies their own AI provider API credentials, pays providers directly for usage, and FlowRunner orchestrates calls across providers without taking a markup on inference. One currency instead of two. The second is whether the human step is in the plan. [Human-in-the-loop](https://flowrunner.ai/concepts/human-in-the-loop) is an execution pattern where AI agents pause autonomously, assemble the relevant context and the decision choices available, route to a human via their preferred channel, and resume the moment the human responds. It is the reason a large share of these workflows exist inside a business at all, and it is priced further apart across this list than anything else on the table. ## The table | Tool | What the meter counts | Allowance and limits on Free | AI on Free | Human-in-the-loop on Free | Unlimited users | First paid rung | | --- | --- | --- | --- | --- | --- | --- | | **FlowRunner** | Execution: one whole run, any number of steps | 100 executions a month, 1 concurrent run, 24 hours of log history, no card, no time limit | Agents on your own keys, no AI meter | Included: email, Slack, WhatsApp, phone, Form; run waits up to 30 days | Yes | Starter $5, 300 executions | | **Activepieces** | Credit: one run regardless of steps, AI the exception | 100 credits a day, no card | Agents included; AI consumes credits | Not stated on the pricing page | No, single user | Plus $20, 10,000 credits | | **Stack AI** | Run | 500 runs a month, 2 projects, 1 seat | Not stated on the pricing page | Not stated on the pricing page | No, 1 seat | Enterprise, custom quote | | **Zapier** | Task: one successful action step | 100 tasks a month, 15-minute polling | Zapier Agents free, 400 activities, 10 per run, billed separately | Request Approval is Professional and above | No, per-seat | Professional $29.99 monthly, about $19.99 annual, 750 tasks | | **Make** | Credit: one module run over one bundle | 1,000 credits a month, 2 scenarios, 10-minute run cap, 15-minute interval | Make AI Agents start on paid tiers | Human in the Loop app is Enterprise, closed beta | Yes | Core about $12 | | **Pipedream** | Two meters: credit (30 sec compute at 256 MB) and AI tokens | 100 credits a month plus 1M AI tokens, 3 active workflows, 3 connected accounts, 300-second run cap | 1M AI tokens included | Not stated on the pricing page | Not stated on the pricing page | Basic $29 billed annually, 2,000 credits | | **Dify** | Message credit | 200 message credits, 5 apps, 1 member, 50 knowledge documents | AI is the meter | Not stated on the pricing page | No, 1 member | Professional $590 per workspace per year | | **Tines** | Live workflow count | 3 live workflows, one-time $50 AI allowance | One-time $50 allowance, not monthly | Not stated on the pricing page | Yes | Unpublished, quote | | **n8n (self-hosted)** | Execution: one whole run | Unlimited executions, on infrastructure you run | AI Agent node on your own model credentials | Send and Wait for Response node | Self-managed; SSO and RBAC are Enterprise | n8n Cloud Starter 20 euros, 2,500 executions, annual | The nine entries below are grouped by what the meter counts, because that is the fact a reader needs before any other. Inside the groups there is no ranking. FlowRunner is first because this is FlowRunner’s page, which is a more useful thing to tell you than a score. ## Run-metered: FlowRunner, Activepieces, Stack AI ### FlowRunner FlowRunner is a visual platform for building workflows and AI agents where the agent calls a human as a tool and waits for the answer. The Free plan is the full platform rather than a sampler of it: it meters volume instead of capability. - **Free plan includes:** all 2,100+ integrations, AI agents, human-in-the-loop, unlimited users, unlimited workflows, single-flow export, 24 hours of log history, community support, and one Free workspace per account. No card and no time limit. Every new account also gets 14 days of the Professional plan on its first workspace, with audit trails and role-based access control, on any real email address. - **What the meter counts and the monthly allowance:** the execution. One execution is one complete run from start to finish, so a run with two steps and a run with twenty each count once. 100 executions a month, one running at a time. - **AI on the free plan:** included, on your own provider keys. You pay OpenAI, Anthropic, Google, or whoever else directly at your own rates, with no AI meter on any plan. - **Human-in-the-loop on the free plan:** included. The agent calls a human-review flow as a tool, suspends while it runs, reaches the person over email, Slack, WhatsApp, phone, or a Form, and resumes with the answer. The run holds open up to 30 days on Free. - **First paid rung:** Starter, $5 a month for 300 executions or $15 for 3,000, same platform, more volume. The [full ladder](https://flowrunner.ai/pricing) continues from there, and the [Free plan announcement](https://flowrunner.ai/blog/flowrunner-free-plan-and-starter-pricing) has the reasoning behind each rung. - **One grounded limit:** one run at a time, and a 24-hour log history window on Free. ### Activepieces Activepieces is an open-source automation platform with a hosted free plan and a Community Edition you can run yourself. Its unit model is the closest match to FlowRunner’s on this list. - **Free plan includes:** unlimited flows, agents, tables, and MCP, on a single user account, no card. - **What the meter counts and the monthly allowance:** a credit, which Activepieces defines as one run regardless of step count, with AI as the stated exception. 100 credits per day. - **AI on the free plan:** agents are included, and AI consumes credits beyond the run. - **Human-in-the-loop on the free plan:** not stated on the pricing page. - **First paid rung:** Plus, $20 a month, 10,000 credits, up to 5 seats, $0.007 per credit in overage. - **One grounded limit:** the free plan is a single user, and AI draws on the same credit pool as the runs. ### Stack AI Stack AI is a builder for AI applications and agents with a run-metered free tier and a sales-led path above it. - **Free plan includes:** 2 projects, 1 seat, community support. - **What the meter counts and the monthly allowance:** runs. 500 a month. - **AI on the free plan:** not stated on the pricing page. - **Human-in-the-loop on the free plan:** not stated on the pricing page. - **First paid rung:** Enterprise, custom pricing. - **One grounded limit:** the next step up is a sales conversation. There is no self-serve paid rung to graduate into. ## Step-, module-, and compute-metered: Zapier, Make, Pipedream This is where the headline number stops meaning what a reader assumes it means, because the count no longer tracks the work. It tracks the shape of the work. ### Zapier Zapier connects apps through Zaps, which run as a linear sequence of steps. Its free plan is the most widely recognized in the category and its unit is the action step. - **Free plan includes:** 100 tasks a month, 15-minute polling, per-seat users. - **What the meter counts and the monthly allowance:** a task, which Zapier defines as one successful action step. A five-step Zap uses five tasks per run, so 100 tasks is 20 runs of that Zap. FlowRunner’s Free plan counts 100 executions, where one whole run is one execution whatever its step count; Zapier’s Free plan counts 100 tasks, where each action step counts on its own. The numbers match and the units do not, which is the single most useful sentence on this page. - **AI on the free plan:** Zapier Agents has its own free allowance of 400 activities a month, capped at 10 activities per run, metered separately from Zapier tasks. - **Human-in-the-loop on the free plan:** the Request Approval action sits on Professional and above. - **First paid rung:** Professional, $29.99 a month billed monthly or about $19.99 on annual, 750 tasks, as of July 2026. - **One grounded limit:** per-step metering, so every branch and enrichment you add changes what an existing workflow costs to run. Side by side on the rest of it: [FlowRunner vs Zapier](https://flowrunner.ai/compare/flowrunner-vs-zapier). ### Make Make builds scenarios from modules on a visual canvas, and it counts both the modules and the data moving through them. - **Free plan includes:** 2 active scenarios, a 15-minute minimum interval, a 10-minute cap on a single run, and unlimited users. - **What the meter counts and the monthly allowance:** a credit, which is one module run over one bundle of data. Make’s own documentation puts a three-module scenario processing ten items at roughly 31 credits. 1,000 credits a month. - **AI on the free plan:** Make AI Agents start on the paid tiers. - **Human-in-the-loop on the free plan:** the Human in the Loop app is Enterprise-only and in closed beta. - **First paid rung:** Core, about $12 a month as of July 2026. - **One grounded limit:** two scenarios, and the 10-minute ceiling on any single run. More at [FlowRunner vs Make](https://flowrunner.ai/compare/flowrunner-vs-make). ### Pipedream Pipedream is a code-forward platform where workflows are steps of hosted JavaScript or Python alongside prebuilt actions. It meters what a server actually spends. - **Free plan includes:** 3 active workflows, 3 connected accounts, unlimited workflow testing, a 300-second execution cap, and a 5-minute minimum schedule interval. No card. - **What the meter counts and the monthly allowance:** two things at once. Compute credits, where one credit is 30 seconds of compute at 256 MB, up to 100 a month; and AI tokens, 1M a month. - **AI on the free plan:** the 1M AI tokens are the allowance, on their own meter. - **Human-in-the-loop on the free plan:** not stated on the pricing page. - **First paid rung:** Basic, $29 a month billed annually, 2,000 credits and 20M AI tokens. - **One grounded limit:** three active workflows, and control flow operators for branching are listed on the Advanced plan at $49. ## AI-message-metered: Dify ### Dify Dify is an open-source platform for building LLM applications, with a hosted Sandbox tier and a Community Edition under the Dify Open Source License. The meter is the model conversation itself. - **Free plan includes:** 1 workspace, 1 member, 5 apps, 50 knowledge documents. - **What the meter counts and the monthly allowance:** message credits. 200 of them. - **AI on the free plan:** AI is the meter. Every message the app produces draws down the same 200. - **Human-in-the-loop on the free plan:** not stated on the pricing page. - **First paid rung:** Professional, $590 per workspace per year, 5,000 message credits a month, 3 members. - **One grounded limit:** 200 messages and a single member. ## Workflow-count-metered: Tines ### Tines Tines runs automation for security and IT teams and positions agents inside governed workflows. Its free edition counts neither runs nor steps. It counts how many workflows are live. - **Free plan includes:** unlimited users, unlimited spaces, unlimited connectors, and a one-time $50 AI allowance. - **What the meter counts and the monthly allowance:** live workflows. Three of them at a time. - **AI on the free plan:** a one-time $50 allowance rather than a monthly one. - **Human-in-the-loop on the free plan:** not stated on the pricing page. - **First paid rung:** unpublished. Paid pricing is a quote. - **One grounded limit:** three live workflows, and an AI allowance that does not refill. ## Self-hosted: n8n ### n8n n8n is a developer-oriented automation platform distributed under a fair-code Sustainable Use License, with a hosted Cloud product and a Community edition you install yourself. Its unit is the same as FlowRunner’s: one execution is one whole run. - **Free plan includes:** the Community edition, self-managed, with unlimited executions on infrastructure you provide. n8n Cloud has no permanent free plan. - **What the meter counts and the monthly allowance:** the execution, one complete workflow run, and on Community there is no cap on how many. - **AI on the free plan:** the AI Agent node, running on model credentials you supply. - **Human-in-the-loop on the free plan:** the Send and Wait for Response node. - **First paid rung:** n8n Cloud Starter, 20 euros a month for 2,500 executions on annual billing. - **One grounded limit:** nothing is hosted for you. The $0 is the license. The server, the upgrades, the patching, the backups, the uptime, and whatever compliance posture your industry asks for are all yours to run and yours to pay for, and that bill arrives monthly whether or not anyone sends you an invoice. The rest of the head-to-head: [FlowRunner vs n8n](https://flowrunner.ai/compare/flowrunner-vs-n8n). ## The names that no longer have a free plan Six vendors a reader would reasonably expect on this list are not on it, and the honest reason is worth more than a padded entry. - **Gumloop.** A 14-day trial rather than a free plan. Pro starts at $37 a month with 20,000 credits, supports your own keys, and adds an 8% orchestration fee. The site now labels workflows “Legacy” and sells agents. - **Lindy.** A 7-day trial. Plus is $29.99 per user per month with 3,000 credits. - **Relevance AI.** The free plan is retired and closed to new signups. Pro is $19 a month on annual billing. - **Microsoft Copilot Studio.** Trial only, and agents built on the trial cannot be published. The standalone plan is $200 a month for 25,000 Copilot Credits. - **Workato.** Quote-based pricing with a trial of unpublished length. - **Relay.app.** Shut down on September 14, 2026 for paying customers, after closing free accounts on August 15 and stopping signups on July 16. If you are holding a Relay export, [this is the rebuild guide](https://flowrunner.ai/blog/relay-app-shut-down-rebuild-your-workflows-free). A free plan that disappeared is a fact about the category, not an accident. Free tiers are expensive to run, and the ones that survive are the ones the vendor decided to fund on purpose. ## What the meter is actually telling you A meter is not a billing detail. It is a claim about what one unit of work is. A step meter says the unit of work is a single action inside a single system: write the row, send the message, update the record. A run meter says the unit of work is the whole errand from trigger to finish. That distinction is philosophical right up until the moment it lands on your bill, because it decides whether adding a branch to a working automation is a design choice or a purchase. The errand is the honest unit, because the errand almost never fits inside one system. The record starts in the tool that created it. The rule that governs it was written somewhere else. The person who has to look at the odd one out is in a third place entirely, on a phone, and the evidence of what they decided has to survive in a fourth. No system of record owns that stretch of ground, and none of them will: each is built to be authoritative about its own data and indifferent to everything past its edge. That stretch belongs to a category of its own: an orchestration layer that sits above the systems and holds a single run open across all four of them. A layer like that treats the person in the middle as a participant with a question to answer, not an outage to route around. [Orchestration as a Service](https://flowrunner.ai/concepts/orchestration-as-a-service) is the name for that category, a platform category that coordinates, governs, and supervises multi-agent environments while keeping humans in control of the decisions that require judgment. FlowRunner is built for that layer, which is why its free number counts errands. ## Four questions to ask any free plan Skip the ranking. Ask these four, in this order, of whatever is in front of you. 1. **What does the number count?** Runs, steps, modules, compute seconds, or messages. Until you know, the number is decoration. 2. **Is AI on its own meter?** Credits that bundle the model cost are a second currency. Your own provider keys are one. 3. **Is the human step in the plan, or above it?** Find the tier where approvals live before you build anything that needs one. 4. **How many people can be in it?** A single-seat free plan is a personal tool. Check the seat count before you design a process two people share. Four answers and the field sorts itself, without a single star rating. ## Start on the Free plan Nine free plans, five incompatible units, and one question that cuts through all of it: what does the number count? On FlowRunner it counts whole runs, 100 of them a month, with all 2,100+ integrations, unlimited users, agents on your own keys, and the human step included, for $0 with no card and no clock. Open a [free workspace](https://app.flowrunner.ai) and build the workflow that has a person in it first. That is the one the meter was always going to argue with, and it is the one worth knowing the answer about before you commit. If you are moving from one of the plans above and would rather walk it through with someone, [send us a note](https://flowrunner.ai/contact); that help is free. ## Frequently asked questions ### Is there a free workflow automation tool with AI agents included? Yes. FlowRunner's Free plan runs AI agents on your own provider keys with no AI meter, Activepieces includes agents on its free plan with AI consuming credits, Dify meters AI directly as message credits, and Zapier Agents is free at 400 activities a month billed separately from Zapier tasks. Verified September 15, 2026. ### Which free automation plans include human-in-the-loop? FlowRunner includes it on every plan including Free, over email, Slack, WhatsApp, phone, or a Form, with the run holding open up to 30 days. Zapier's Request Approval action is on Professional and above. Make's Human in the Loop app is Enterprise and in closed beta. n8n's Send and Wait for Response node ships with the self-hosted Community edition. ### What is the difference between a task, a credit, and an execution? An execution is one complete workflow run, however many steps it contains. A Zapier task is one successful action step, so a five-step Zap uses five tasks per run. A credit means different things by vendor: one run on Activepieces, one module run over one bundle of data on Make, and 30 seconds of compute at 256 MB on Pipedream. ### Is n8n free? The self-hosted Community edition is free under a fair-code Sustainable Use License and runs unlimited executions on infrastructure you provide, which means the server, the upgrades, the patching, and the backups are yours. n8n Cloud has no permanent free plan; its first rung is Starter at 20 euros a month for 2,500 executions on annual billing. ### Does FlowRunner's Free plan expire? No. It is $0 with no card and no time limit: 100 executions a month, one run at a time, 24 hours of log history, unlimited users, unlimited workflows, and all 2,100+ integrations. Every new account also gets 14 days of the Professional plan on its first workspace, and when that ends the workspace moves to Free with flows still running. By [Mark Piller](https://flowrunner.ai/about), Founder of FlowRunner [Editorial policy](https://flowrunner.ai/editorial-policy) --- Markdown version of https://flowrunner.ai/blog/best-free-ai-workflow-automation-tools. Site index: https://flowrunner.ai/llms.txt