Back to Blog

How to protect free runs in a workflow automation tool

Free workflow runs abuse puts your best customers at the back of the queue. What a free run costs, how to meter it and which basic fixes break.

It's 1:47 p.m. on Wednesday, six days after you shipped the run meter, and your biggest customer has just written in: "Our lead-routing workflow fired at 1:00 and still hasn't started. It was an hour late yesterday too. We pay you $399 a month for this."

You were proud of that meter. It took three weeks, every free account now stops at 100 runs a month, and the day it went live you posted a screenshot in your team channel: zero accounts over the limit. You open the same dashboard now and it still says zero. Then you open the queue, and just over 7,000 runs are waiting in it.

That's free workflow runs abuse from the inside. Nothing is over its limit, and the customers paying you the most are the ones waiting.

This is a scenario, so check the arithmetic yourself. Say 60 people each run 50 free accounts, which makes 3,000 accounts. Each account fires one scheduled run at 9, 10, 11, 12 and 1, every working day. Five runs a day for 20 working days is exactly 100 runs a month, so your meter never has a reason to object.

Capacity:            50 workers x 120 = 6,000 runs an hour
Paying customers:    3,600 runs an hour
Honest free users:     900 runs an hour
The 3,000 accounts:  3,000 runs an hour
Total demand:        7,500 runs an hour
Queue grows by:      1,500 runs an hour

By 2 p.m., 7,500 runs are waiting. At 6,000 runs an hour, that's a 75-minute delay for anything joining the line. Once those scheduled runs stop arriving you clear 1,500 an hour, so the queue empties at 7 p.m. (7,500 ÷ 1,500 = 5 hours). Tomorrow morning it starts again.

Illustrative queue delay02040608091011121234567Minutes of delayHour, 9 a.m. to 7 p.m.
Illustrative queue delay0204060809111357Minutes of delayHour, 9 a.m. to 7 p.m.

Your heaviest users feel this most, because they have the most runs in the line. A lead-routing flow that starts 75 minutes late means a sales team calling a lead who's gone cold, and the person who gets blamed inside that company is whoever picked your tool.

The bill comes next. In this scenario a run costs you 2 cents, so those accounts cost 3,000 × 100 × $0.02 = $6,000 a month. Your 180 paying customers average $79 a month, which is $14,220. About 42 cents of every dollar they pay you now goes to running automations for 60 people who'll never pay. With $120,000 in the bank and $20,000 a month going out, you've got 6 months. Without that $6,000 you'd have 8.6 ($120,000 ÷ $14,000).

You refund the $399 customer's month. Four smaller customers ask for a credit at $79 each, another $316. If the big account leaves anyway, $4,788 a year goes with it.

So you open a blank doc and title it "Cut the free plan?" Nearly all of your 180 customers started there. Drop it to 25 runs and the same 300,000 runs would take 12,000 accounts instead of 3,000 (300,000 ÷ 25), while someone building their first real workflow burns through 25 test runs before lunch and never comes back. The free tier you were proud of is now the thing you're being pushed to cut, because of 60 people you can't even name.

Why a per-account limit doesn't protect free workflow runs

Your meter counts runs inside one account. An account costs nothing to create, so each new one arrives with a fresh 100, and the limit you built only decides how many accounts a person needs.

Workflow tools make that trade attractive. A free run keeps working when nobody's watching, so someone can build an automation once, set a schedule and walk away. A free plan like yours can end up powering someone's paid automation service for their own clients, with your workers doing the work.

Here's the scenario's 2 cents, broken down, and the place in line each run takes.

Free accountcreatedSchedule firesat 9 a.m.Run joinsthe queueWorker time0.4 centsModel step1.3 centsLogs and retries0.3 centsPaying customerswait behind itIllustrative total2 cents a runYour monthly billgrows
Free accountcreatedSchedule firesat 9 a.m.Run joinsthe queueWorker time0.4 centsModel step1.3 centsLogs and retries0.3 centsPaying customerswait behind itIllustrative total2 cents a runYour monthly billgrows

How to protect free workflow runs from the inside

Keep the meter, and add the changes that protect your capacity, because they work no matter who's behind the accounts.

Give free runs their own queue. Do this first. Split the 50 workers into 40 for paying customers and 10 for the free plan. Paid capacity becomes 4,800 runs an hour against 3,600 of demand, so the line your $399 customer sits in is empty again today.

Meter what a run costs. A two-step run and a 400-pass loop both count as one run, and they don't cost the same. Count steps or worker seconds instead, and show that count in the product.

Cap each free run. Set a maximum step count and a time limit, switch off parallel branches, and count retries against the allowance.

Spread out free schedules. Limit them to once an hour at most, add a random offset so they don't all land on the hour, and pause schedules on accounts nobody has opened in 14 days.

Set a hard daily budget for the free plan. In the scenario, the 3,000 accounts cost $300 a day (15,000 runs × $0.02). When the budget's spent, free runs wait for tomorrow and paying customers don't notice.

Your paying customers are safe now, and the problem has moved. Free demand is still 3,900 runs an hour (900 + 3,000) against 1,200 of capacity, so the free line grows by 2,700 an hour. By 2 p.m. it holds 13,500 runs, which is more than 11 hours of work (13,500 ÷ 1,200 = 11.25). Someone trying your product for the first time at 2 p.m. sees their first run start after 1 a.m. That's the person your free plan exists for.

The basic checks on new accounts, and how each one breaks

So the next job is keeping repeat accounts from getting the runs at all. Nearly everyone tries these first, and each one catches something.

Email verification and unique emails. Verification clears out typos and dead addresses, and a unique-email rule stops the same address signing up twice. Keep both. Neither notices a second real mailbox, because verification only proves someone can open an inbox, and now and then a real person's code lands in spam.

Disposable-domain blocks. A blocklist stops the throwaway services on it. It misses new domains and ordinary mailboxes, it can turn away people using privacy forwarding, and someone has to keep it updated forever.

Gmail dots and plus addressing. Gmail ignores dots, so jane.doe@gmail.com and janedoe@gmail.com are one inbox, and mail to jane+runs@gmail.com reaches Jane too. A unique-email rule sees different people. Accounting for that catches one inbox written several ways and no more. It's also one more rule someone has to keep up to date.

CAPTCHA. It filters simple scripts. It helps less here, because a person can pass it by hand, and one human moment at signup buys 100 unattended runs a month. Meanwhile you're paying for it in fees, lost signups and people using screen readers.

Phone or card requirements. These stop casual second accounts, and anyone with another number or card gets through. You pay for every text, you lose people without a reliable mobile number, and a card wall puts off the developer who wanted to try one automation first.

One account per IP address and VPN blocks. These catch repeat signups on one connection and known VPN ranges, and they miss every other network. The wrong people get caught fast in this market: an agency building automations for 12 clients from one office, or a company whose whole team sits on a corporate VPN. Each one becomes a ticket and an exception.

Rate limits and usage caps. A limit of 10 runs a minute stops one account going wild. It can't see 3,000 accounts firing once an hour each, and every new account gets a fresh cap. Set it too tight and real users hit it in a busy hour.

Manual review. A person can spot things a rule can't. But 3,000 accounts at one minute each is 50 hours, next month's batch needs the same 50, and a tired reviewer suspends the agency along with everyone else.

Add it all up and your form asks for a verified email, a puzzle, a phone number and a card before anyone's built a single workflow. The 3,000 accounts you already have carry on regardless, since a signup check only runs at signup.

Treat free workflow runs as something you grant

Every check above asks about the signup. What you need to know comes a moment later: has this person already had this allowance?

So treat the runs as a grant. Put the signup allowance and each monthly reset behind one server function, start no schedule until that function says yes, and write down what you're giving each person, with room for agencies and teams. Decide what a no looks like too: a paid plan and a way to reach you, since some of the people you turn away will be real.

Your database can tell you whether account 48,211 got its runs. It can't tell you who's behind account 48,212.

Get a clear answer on every free allowance

AbuseZero exists to evaluate whether this claim looks like a repeat, and it starts where each fix above gave out. A blocklist knows domains, an IP limit knows one connection and your meter knows one account. You get multi-layer AI risk analysis that examines 140,000+ signals, detects linked claims and flags abuse that single checks miss, reading those signals together and never one at a time. There's also a proprietary algorithm, with AI that checks the claim again.

New mailbox every time? AbuseZero recognizes repeat devices across new accounts, even when email addresses keep changing, and it checks mailbox aliases, disposable domains and mail routing as well. Where an IP limit sees one address, AbuseZero sees beyond it, with VPN, hosting, network reputation, Tor and proxy intelligence. It blocks spammy bots, browser automation and malicious scripts while allowing genuine AI agents and assistants. Creating a fresh account isn't enough to erase previous activity, which is something no per-account meter can say.

Your server makes one API call before the runs are granted and gets a clear allow or deny, under rules you control. Eligibility can be once per user for the signup allowance and once per calendar month for the reset. A trusted agency can go on an allow list. Observe mode shows you what would be denied before anyone is blocked, and the quickstart gets your first benefit protected. You don't have to build your own abuse prevention system, and your meter, queues and caps stay where they are.

Frequently asked questions

What is free workflow runs abuse?

It's one person collecting more free runs than you meant to give them, usually by running several accounts. Each account can stay inside its limit the whole time.

Should I remove my free plan to stop this?

Keep that for last. Separate the queues and cap each run first, then read should you remove your free tier before you decide.

Do I still need a meter if I evaluate each claim?

Yes. AbuseZero's evaluation decides whether a claim gets the allowance, and your meter decides when it runs out.

Keep the free plan and the fast queue

Keep reading

Copying your prompt…

Copied the to your clipboard. Give it to your coding agent & it will handle the rest.