What Zapier Agents' free tier actually gives you
What the free plan actually allows: what counts as an activity, why test runs cost you, and when the allowance resets. Checked against Zapier's own documentation.
The number that matters is not 400
Zapier Agents' free plan gives you 400 activities a month. That is the figure usually quoted, and it is the one least likely to cause you a problem.
The limit that actually interrupts you is the one on a single run: on the free plan an agent can use up to 10 activities before it stops and asks you for input rather than finishing the job. On the Pro plan that ceiling is 40.
You can have 380 activities left for the month and still have a run stop halfway. The two limits are unrelated, and only one of them tends to show up mid-task.
So what is an activity?
Zapier counts an activity each time your agent does one of these things:
- Uses a trigger
- Answers from a knowledge source
- Runs an action
- Makes a web browsing or scraping request
- Runs a web search
- Sends a message via the Chrome extension
The useful shape to hold in your head: an activity is the agent reaching outside itself. Searching, fetching, doing. That is what gets counted.
Chatting with the agent is not billed. You can talk to it as much as you like.
On the free plan, test runs count
This is the one that catches people, and it is worth knowing before you start rather than after.
On the free plan your test runs draw down the same monthly allowance as everything else. On Pro they do not. So an afternoon of building and re-building an agent is spending real allowance every time you try it, even though nothing has gone into production.
It is not a reason to avoid testing — testing is how you get an agent that works. It is a reason to make each test deliberate rather than mashing the button and seeing what happens.
The allowance resets on your billing date
Not on the first of the month. Your 400 refreshes on the same date each month that your account uses for billing, which for most people is whenever they signed up.
Worth checking once, because "it resets next month" and "it resets on the 9th" are quite different pieces of information when you are three days from running out.
Agent activities do not touch your Zap tasks
Zaps and Agents are two different products, and they are metered separately. Running your agent does not consume the task allowance your Zaps draw on.
This matters mostly as reassurance: people sometimes avoid trying Agents because they are protecting a task budget that Agents was never going to touch.
It is a permanent free plan, not a trial
New Zapier accounts get a 14-day Pro trial. That trial covers Zaps, premium apps and multi-step Zaps — Zapier's own trial documentation does not mention Agents at all.
Agents on the free plan is listed as a feature of that plan, not as something borrowed from the trial. So the free tier described here is not a countdown you are quietly running down.
One caveat we would rather state than skip: we have not tested whether a brand-new account inside its trial window displays the Pro per-run ceiling of 40 instead of 10. If you are in your first fortnight and the numbers you see differ from the ones here, that is the likeliest explanation.
What running out actually looks like
If you use all 400 in a month, the agent does not disappear. It still talks to you — but it stops doing. Actions, web search, browsing, triggers and knowledge lookups all stop working until the allowance resets.
An agent that answers but cannot act is a confusing thing to be handed with no explanation, so it is worth recognising.
The number we cannot give you
We deliberately have not told you what a typical research task costs in activities, because we have not measured it and we are not going to guess at it. It depends on how many pages the agent decides to read, which depends on how you have written its instructions.
What we can say is the mechanism: bound what the agent is allowed to read and you bound what it spends. Precise instructions are cheaper than vague ones, and that is as true of your allowance as it is of the quality of the output.
If your agent has already stopped mid-run and asked you something, we wrote about what that means.
Where these numbers came from
Checked on 23 August 2026 against Zapier's usage documentation and its pricing page. Platform limits change, so treat that date as part of the fact — if you are reading this much later, verify before relying on it.
Build your first agent in 30 minutes
One guided session, and you finish with a working agent you can run and adapt. No coding experience required.