Zapier Agents and Make AI Agents both let a small team give an AI model instructions, tools and knowledge and have it work inside an automation platform, but they bill differently. Zapier measures agent work in activities, which are separate from Zap tasks and capped by plan. Make bills everything in credits, with AI usage varying by tokens. The cheaper option depends on how many runs you need and how many steps each run takes, so the useful skill is estimating your own usage, which this guide walks through.
This comparison is for owners and operations leads at small businesses who are choosing between the two platforms, or who already use one and want to understand the cost of adding agents. You will finish with a side-by-side feature table, the documented limits and overage rules, and a worksheet for estimating monthly cost. Facts and prices were checked on Zapier and Make documentation on 28 October 2026. Both vendors change plans and rates, so treat the numbers as a snapshot and confirm them on the live pages.
What each product is
Zapier Agents are AI agents you configure in Zapier. They use triggers, actions, knowledge sources, web browsing and search. Zapier's help documentation counts agent work as activities, such as an agent using a trigger, running an action, answering a question from a knowledge source, a web search or a browsing request.
Make AI Agents are agents you create in Make with a connection, a name, a model and a system prompt. Make's documentation describes tools you can attach (modules, scenarios and MCP servers) and knowledge files that give background information. Agents run inside Make's scenario model, which is why credits are the billing unit.
If you are choosing a platform more broadly rather than just for agents, see n8n vs Make vs Zapier: Choosing an Automation Platform.
Feature comparison
| Area | Zapier Agents | Make AI Agents |
|---|---|---|
| Setup | Configure an agent with instructions, tools and knowledge | Set a connection, name, model and system prompt, then add tools and knowledge |
| Tools | Zapier triggers and actions, web browsing, web search, knowledge sources | Modules, scenarios and MCP tools |
| Knowledge files | Knowledge sources | Upload TXT, PDF, DOCX, CSV, MD and JSON; 20 MB per file and 20 files per agent in the documentation for the earlier version of the app |
| Model providers | Not stated in the sources checked | All plans can use Make's AI Provider; paid plans can connect custom providers such as OpenAI or Anthropic Claude |
| Sharing | Agent sharing is not available on Free or Pro; Team and Enterprise share one activity pool | Agents are shared across all team members, like connections |
| Human approval | Not covered in the sources checked; check Zapier's documentation for approval options on your plan | Design approval steps in your scenario |
The Make knowledge limits come from a page that says it describes a previous version of Make AI Agents and links to a newer one, so check the current documentation before depending on them.
How Zapier bills agents
Zapier Agents are an add-on, billed separately from your task allowance. The help article gives these plan limits:
| Plan | Agent activities per month | Per-run cap | Notes |
|---|---|---|---|
| Free | 400 | 10 activities | Tests count toward the quota |
| Pro | 1,500 | 40 activities | Testing an agent does not count |
| Enterprise | Custom | 40 activities | Agent sharing available |
On Team and Enterprise accounts, all users share one activity pool. The count resets on your billing date, which is the day you signed up or last upgraded. Activities do not affect your Zapier task usage. Zapier's pricing documentation lists the Agents add-on starting at about USD 33.33 a month on annual billing for Pro, billed separately from tasks. These prices were stated on the pricing page at the time of writing and can change.
What counts as an activity
Each of these counts as one activity: an agent using a trigger, answering from a knowledge source, running an action, a web browsing or scraping request, a web search, or a message sent to the agent through the Chrome extension. An agent that searches the web, reads a page and sends a message may therefore use several activities for one job. The per-run cap stops one run from using the whole month's quota, and if a run hits the cap with work left, the agent asks for your input.
What happens at the limit
Agents stop running, including actions, browsing, searches, triggers, knowledge lookups and Chrome extension access. You can still chat with the agent. Zapier emails you at 80% and 100% of the limit. When you upgrade on your billing date, the limit resets and you are charged the full amount; on other days you are charged a prorated amount and the limit increases without resetting.
Zap tasks still matter
The Zaps that your agent triggers, or that run alongside it, still count as tasks. Zapier's pricing page lists Pro from USD 19.99 a month on annual billing (USD 29.99 monthly) for 750 tasks, and Team from USD 69 a month on annual billing (USD 103.50 monthly) for 2,000 tasks, with higher tiers up to 2 million tasks. Annual billing is about 33% off the monthly rate. If pay-per-task is turned on, overage is billed at 1.25 times the base rate on annual plans or 2.5 times on monthly plans, and usage stops at three times your subscription. These figures are from zapier.com/pricing at the time of writing; verify before you buy.
How Make bills agents
Make's billing unit is the credit. The credits documentation says:
- Non-AI apps use 1 credit per operation, and third-party AI apps also use 1 credit per operation.
- Make's AI Provider (available on all plans) uses credits depending on tokens and operations.
- With a custom AI provider connection on a paid plan, you pay Make per operation and pay your provider directly for tokens.
- With the automatic AI provider connection, Make chooses the provider, and credits depend on tokens, operations and other usage factors.
- Usage is fixed (a set number per run) or dynamic (varying with tokens, file size, pages or run time).
The Free plan includes 1,000 credits per month, according to the pricing page. Make's extra-credits documentation uses an example of a USD 9 plan with 10,000 credits, which works out to USD 0.0009 per credit. Current plan prices and paid-plan credit allowances are not reproduced here because they could not be confirmed from Make's own pricing table; check the live pricing page before you compare costs.
What happens at the limit
When credits run out, scenarios stop until more credits are added, and incoming webhooks queue up to your queue limit. Paid plans can buy extra credits: they cost the plan's per-credit rate plus a 25% surcharge, are bought in 1,000-credit increments manually, and can be auto-purchased in 10,000-credit blocks when subscription credits run out. Auto-purchases are capped per cycle at your subscription's credit amount. Included credits are used first. Unused credits expire at the end of the term.
For how Make counts operations in detail, see Make.com Operations and Credits Pricing Explained.
Compare cost with your own numbers
Vendor prices change, but the method does not. Use this worksheet.
- List the jobs. For example: triage support emails, qualify new leads, summarise form submissions.
- Estimate runs per month for each job. Use your real volumes, such as 300 emails.
- Count the steps per run.
- For Zapier Agents, count activities: triggers, actions, searches, knowledge lookups and browsing.
- For Make, count operations (1 credit each for most modules) and add the AI token cost.
- Add surrounding automation. Count Zap tasks for Zapier and credits for any non-agent scenarios in Make.
- Multiply and add a buffer. Multiply runs by steps, add 20 to 30 percent for retries, testing and growth, and compare against plan allowances.
- Check the per-run caps. On Zapier, a run that needs more than 10 activities on Free or 40 on Pro will stop and ask for input.
- Price it. Take the cheapest plan that covers the estimate on each vendor's live pricing page, and note the overage rules.
Run a one-week pilot on real work before committing, and record actual activities or credits used per run. Real usage often differs from the estimate because agents may take extra steps to complete a task.
Choosing: questions to ask
| Question | Points to |
|---|---|
| Do you need the widest catalogue of ready-made app connections in one place? | Check both catalogues for the apps you use |
| Do you want to bring your own model provider and pay tokens directly? | Make paid plans (custom AI provider connection) |
| Do you want predictable agent usage limits separate from task counts? | Zapier activities, with per-run caps |
| Do you want visual control over multi-step scenarios and routing? | Make scenarios |
| Do you need a person to approve before an action? | An approval step built in Make; for Zapier, check its current documentation for approval options |
Neither is automatically cheaper. A job that takes few steps but runs often looks different from a rare job with many tool calls.
Keep agents safe and cheap
- Give agents narrow instructions and few tools. Each extra tool call is an activity or credit.
- Start with read-only actions and add write actions only after testing.
- Add a human approval step before an agent sends external email or changes records.
- Set alerts: Zapier emails at 80% and 100% of the activity limit. In Make, watch credit usage in your organisation dashboard.
- Prefer a plain automation for deterministic jobs; use an agent only where judgement is needed.
Checklist
- Jobs, volumes and steps estimated for both platforms.
- Live pricing pages checked on the day you decide.
- Overage behaviour understood: pay-per-task and activity caps in Zapier; extra credits in Make.
- Pilot run with real data, with usage recorded.
- Approval step in place for any outward-facing action.
References
- Zapier, pricing: https://zapier.com/pricing
- Zapier, pricing details (markdown version): https://zapier.com/pricing.md
- Zapier Help, how Zapier Agents usage is measured: https://help.zapier.com/hc/en-us/articles/26559132765325
- Make, pricing: https://www.make.com/en/pricing
- Make Help, credits: https://help.make.com/credits
- Make Help, extra credits: https://help.make.com/extra-credits
- Make Help, manage AI agents: https://help.make.com/manage-ai-agents
- Make Help, Make AI Agents: https://help.make.com/make-ai-agents
Comments
Questions and tips are welcome. Comments are checked by the EpiFive team before they appear; please do not share passwords, phone numbers or other private details.