What Counts as a Zapier Task (and What Is Free)

A Zapier task is one successful action step — not one step in your Zap, and not one Zap run. Triggers are free. Filter and Paths steps are free. Any step that errors, halts, or never runs because a filter stopped it is free. So are all steps from a set of Zapier's own apps: Formatter, Delay, Looping, Sub-Zap, Digest, Zapier Manager and Storage, plus everything in Zapier Tables and Zapier Forms. That is why a Zap with a trigger, a filter, a Formatter step and two actions bills 2 tasks per run, not 5 — and why the most expensive mistake in Zapier budgeting is counting the boxes on your screen.

The complete list of what bills and what doesn't

Most of this comes from that one help page; check the rows that matter to your build against it before you budget on them. The free column is the one that decides your plan.

Two rows deserve a warning. The free-apps list names the "Sub-Zap app", while the counting list names the "Call a Sub-Zap" and "Return From a Sub-Zap" steps as chargeable. Both sit on the same page; the specific rule is the one to budget against, so assume the call and return steps each cost a task. And error handling is not a discount: a failed step is free, but a successful step inside the error handler you built to catch it bills like any other action.

Search actions turn on a setting you picked when you built the step. Zapier's page: with "Yes (proceed if nothing found)", the search uses 1 task; with "No", it uses no tasks. A search that finds its record is a successful action and bills 1 task either way. The setting only decides the empty result: proceed and pay, or halt free.

Worked example 1: a lead capture Zap

A form submission arrives, a filter drops the junk, Formatter splits the name, the contact goes into the CRM, Slack gets a notification.

Say 400 submissions a month and a filter that passes 60 percent: 400 x 0.60 = 240 runs continue, and 240 x 2 = 480 tasks a month. The 160 stopped runs cost nothing.

Budget the naive way — five steps, 400 submissions — and you get 400 x 5 = 2,000 tasks, which lands on the 2,000-task plan at $49 a month billed yearly, or $49 x 12 = $588 a year. The real 480 fits the 750-task plan at $19.99 a month billed yearly, $19.99 x 12 = $239.88 a year. The gap is $49 - $19.99 = $29.01 a month, $348.12 a year, for a Zap that never changed. Prices from zapier.com/pricing, read 4 September 2026, in USD.

Worked example 2: an order Zap with Paths

Branching is where estimates fall apart. This Zap watches for paid orders, formats the total, looks up the customer, then splits: high-value orders join a VIP list and get announced in Slack, the rest go to the standard list.

At 900 orders a month with 20 percent on Path A: 900 x 0.20 = 180 and 900 x 0.80 = 720. Tasks: (180 x 3) + (720 x 2) = 540 + 1,440 = 1,980 a month. That fits the 2,000-task plan at $49 a month billed yearly, $588 a year — with 20 tasks of headroom, so one busy month or one new action step pushes you over.

Count boxes instead and a Path A run has six steps, so 900 x 6 = 5,400 tasks. That is past the 2,000-task tier, and the next tier up is 10,000 tasks at $129 a month billed yearly, $129 x 12 = $1,548 a year. Difference: $129 - $49 = $80 a month, $80 x 12 = $960 a year. Zapier's slider publishes seventeen priced stops, and every one of them is listed below.

The four things that make the bill bigger than you planned

Replaying a run rebills the steps that already worked. Take a three-action Zap where the third action fails on 150 runs because an API key expired. Those runs each billed 2 tasks (two successes, one free error): 150 x 2 = 300. You fix the key and replay all 150 entire runs, which bill 3 tasks each: 150 x 3 = 450. Total 750 tasks for 150 outcomes — 750 / 150 = 5 tasks per outcome, not the 3 you budgeted. The rule names replaying an entire run: successful steps that execute a second time are billed a second time.

Sub-Zaps bill the wrapper as well as the contents. A parent Zap that calls a sub-Zap containing three actions, then does one more action itself, bills: Call a Sub-Zap (1) + three actions inside (3) + Return From a Sub-Zap (1) + the final parent action (1) = 6 tasks. Both triggers are free. The same four actions built inline bill 4. At 2,000 runs a month: 2,000 x 6 = 12,000 versus 2,000 x 4 = 8,000. Inline fits the 10,000-task tier at $129 a month billed yearly ($1,548 a year). The sub-Zap version does not: 12,000 passes 10,000, and the next slider stop is 20,000, which is $189 a month billed yearly against $129 at 10,000 — $60 a month more, or $720 a year. The ratio is the part that stings: the wrapper adds 4,000 tasks a month, 50% on top of the 8,000 the same four actions cost inline, and it pushes you over a tier boundary you would otherwise have sat inside.

Zapier Lead Router bills 5 tasks per step. One lead through one routing step is 5 tasks. Route 600 leads a month and routing alone is 600 x 5 = 3,000 tasks — before a single action in the Zap around it, and already half again the 2,000-task plan.

Zapier MCP bills 2 tasks per tool call. An assistant making 40 tool calls a day across a 30-day month is 40 x 30 = 1,200 calls, or 1,200 x 2 = 2,400 tasks — the least predictable line in the model, since tool calls are easy to trigger and hard to count afterward.

Products that don't bill at one task per action

The AI row changed on 15 June 2026, and it moves a budget by a factor of three. Zapier now prices those steps by model tier: “Your model tier sets the base cost per run (Standard 1x, Advanced 3x, Premium 5x). Tool calls add to that base cost at the same rate.” The tier that matters is the one you get without choosing: “New steps default to the Advanced tier (3x) to allow you to use tools with your AI by Zapier step.” Connecting your own AI account brings it back to 1×. So a Zap you costed at one task per AI step is, by default, costing three.

There is also a brake worth knowing about before it stops something at 3 a.m.: “If a step reaches 75 tasks in a single run, your Zap pauses and asks you to approve before continuing.”

The Code row used to be the one rate with no number against it, and this article declined to invent one. The rate is published now: Zapier's task usage rates page, read on 28 October 2026, lists extended runtime on its own line as "1 task per 30 seconds beyond plan allowance." The same page adds one line this article did not carry: the SDK, in beta, which it marks "Free in beta." The page also warns that "Rates may change over time," so treat all of this as dated, not settled. One more line, quoted because it says the opposite of what the rest of the page implies: “Zapier Agents usage does not count as tasks. Agents use a separate activity quota.” The size of that quota is not stated there.

The price ladder, in USD

Read from zapier.com/pricing on 4 September 2026 and re-read stop by stop on 28 October 2026. The seven priced stops this article already carried are unchanged to the cent; the other ten appear here for the first time. Team is not sold below 2,000 tasks, which is why its two lowest cells are empty. Free is 100 tasks a month. Above Team sits Enterprise, quoted by sales.

The monthly column is almost exactly 1.5 times the yearly one — $49 x 1.5 = $73.50, $289 x 1.5 = $433.50 — which is the arithmetic behind the advertised "Save 33%", since 1 / 1.5 = 0.667. Run the discount on all 32 published pairs and 31 of them do deliver 33% or better, almost all of them landing on 33.3%. One pair does not: Professional at 500,000 tasks, where $1,499 yearly against $2,199 monthly is a discount of 31.8%. The other end is Professional at 2,000,000 tasks, the most generous on the page at 33.5%. Use the 1.5x relationship as a rough test of list pricing, not as a rule. Note that 100 free tasks is 50 runs of the two-task Zap above — Free is a place to learn the tool, not to run a business on.

How to estimate your month before you buy

Five lines of arithmetic per Zap. First, cross out every free step: the trigger, every Filter, every Paths step, and every step belonging to Formatter, Delay, Looping, Sub-Zap, Digest, Zapier Manager, Storage, Zapier Tables or Zapier Forms. Second, count what is left, separately for each path — a branching Zap costs a different amount on each branch. Third, estimate how many runs reach each branch per month: trigger volume times filter pass rate times each path's share. Fourth, multiply and add across branches and Zaps.

Fifth, add the specials: 5 tasks per Lead Router step, 2 per MCP tool call, 1 per Code by Zapier step, and 1 more for every 30 seconds a code step runs beyond the plan allowance. Then add a margin for replays. No average replay rate is published, so this article will not supply one — pull your own Zap History for the last full month, count the replays, and price each at the full per-run cost, not the cost of the failed step alone.

Where to check what you actually used

Zapier names two places on its own help page: the Task Usage page and Zap History. Task Usage tells you how much of the plan you have consumed; Zap History tells you which Zap consumed it, the one you need when the number surprises you. Compare your estimate against a full month of Task Usage. If they are far apart, the gap is almost always one of four things: a search step set to proceed on empty results, replays after a broken integration, a sub-Zap paying two tasks of overhead per run, or Lead Router at 5 per lead. What happens when you hit your limit, and how overages behave, is not on the help page read for this article, so it is not stated here — check it on your own plan first.