Article

Budget for Zapier Next Gen and SDK billing on December 1

Automation owners can estimate December 1 task usage, account for AI and Code charges, and decide whether to pay overages or buy more capacity.

Editorial illustration for Budget for Zapier Next Gen and SDK billing on December 1: a controlled task flows from input to output. Not documentary evidence.

Zapier will start counting Next Gen Zaps and Zapier SDK usage against plan tasks on December 1, 2026, at 17:00 UTC (noon ET). Automation owners should budget for successful billable operations, not just workflow runs. Extra cash charges depend on whether total account usage exceeds the included allowance.

Who needs to prepare for December 1?

Next Gen Zaps are workflows that AI can help build and that Zapier then hosts. Access remains limited and by request on Professional, Team and Enterprise—not Free. Team requests require an owner or administrator; Enterprise customers must contact Support. The temporary free usage period does not make the underlying subscription free.

Most Classic Zaps are unaffected. The exception is SDK usage inside Classic Code by Zapier steps. The SDK is a developer toolkit for using connected apps from code; its Code-step integration is JavaScript-only, in open beta and available across plans, with Enterprise opt-in. That broader SDK eligibility does not grant Next Gen access.

There is no need to import an ordinary Classic Zap just because billing changes. Importing creates a copy, and the original keeps running unless turned off. Before enabling a replacement, check whether both versions would do the same work.

Which operations consume tasks?

Under Zapier’s task-accounting rules, an ordinary successful app action uses one task. Starting triggers, filters, formatting and branching are free. Looping itself is free, but billable actions repeated inside a loop still count. Zapier Tables and Forms actions are also free.

For code-based work, the dedicated billing guide adds these distinctions:

  • An external call through Zapier, using a connected account, normally uses one task. This is not a blanket charge on every internet request made by code.

  • SDK operations within Next Gen count within that workflow; there is no second SDK charge. Outside Zapier, successful SDK actions and external calls consume tasks directly.

  • Batch items count separately. Checking an SDK trigger inbox is free, but leasing an event to process it costs one task. That is distinct from a free hosted starting trigger.

There is a documentation mismatch for logic-only runs. The task-usage help page says a Next Gen run doing only filtering, formatting, loops or branching can use zero tasks; the product FAQ says a completed run uses at least one. Confirm this edge case with Zapier before relying on it at scale. The examples below contain billable actions and do not depend on that disputed minimum.

How many tasks would 100 runs use?

These are illustrative calculations for work after the cutover, not measured bills. Assume 100 successful runs, no retries or replays, no build-time calls and no unrelated account activity.

Ordinary actions: three separately executed, one-task actions per run use 100 × 3 = 300 tasks. Count the actions actually reached: branches and loops change that number. A failed action does not consume a task, but successful work elsewhere in a failed workflow still can. Replaying successful Classic steps can charge for them again.

AI-assisted work: AI by Zapier’s tier multipliers are Standard 1, Advanced 3 and Premium 5; new steps default to Premium. Its formula is multiplier × (1 + successful tool calls). Standard does not support tools. Using your own AI-provider account has a multiplier of 1, with tools restricted to eligible Advanced or Premium models; provider charges remain a separate budget item.

Add one Premium AI step making exactly two successful tool calls to each of the 100 runs. Assume those tools are separate from the three ordinary actions already counted. The AI step uses 5 × (1 + 2) = 15 tasks; total usage becomes 100 × (3 + 15) = 1,800 tasks. Do not count the AI step’s tools a second time.

From December 1, AI by Zapier uses that pricing through MCP and the SDK too. It is not universally a one-task SDK action or a two-task MCP call. In a Zap workflow, the AI-step limit defaults to 75 tasks, can rise to 500, and pauses the step for approval when reached; it is not an account-wide cash cap.

Classic Code: base, runtime and SDK charges stack. On Professional or Team, assume each successful Code step runs for 35 seconds and performs three ordinary, successful SDK actions. Each run uses 1 base task + 3 SDK tasks + 1 runtime task = 5 tasks, or 500 tasks for 100 runs. The extra five seconds exceed the 30-second standard runtime and round up to one additional block. Enterprise’s standard runtime is two minutes; Free has one second and no extended runtime.

How do task counts turn into a bill?

Find your overage rate in Billing settings: it depends on the plan and billing cycle. For an illustrative account with 750 included tasks, 1,800 total tasks and overages enabled, the excess is 1,050 tasks. At rate r, the extra charge is 1,050 × r—not 1,800 × r. Subscription fees, taxes and external services are separate.

The overage ceiling is three times the allowance in total: 2,250 tasks for a 750-task plan. Workflows stop there. Without overages, new Zap runs are held at the included allowance until the billing-cycle reset, unless you change capacity.

Professional and Team overage settings require the owner. Enterprise accounts created after June 2025 must contact their account team to disable overages. Disabling them does not cancel accrued charges.

For the buying decision, add existing account consumption to the new forecast first. Spare included capacity may absorb the change without an immediate overage bill, but leaves less room for other work. For sustained excess, compare the overage estimate with a larger allowance on the same billing term. Do not compare an annual-plan monthly equivalent with a month-to-month quote.

What costs sit outside the running workflow?

An AI client can execute app actions while building a workflow. The getting-started guide says those actions can already consume Zapier MCP tasks, and the client may separately charge for AI tokens. A free runtime offer is not a promise that the entire build costs nothing.

MCP—Model Context Protocol—connects AI clients to tools. Its ordinary successful tool-call rate is two tasks, drawn from the shared allowance. From December 1, the named exceptions make free By Zapier apps unmetered through MCP and apply AI by Zapier’s own rate. Do not multiply every call indiscriminately by two.

For workflows that also collect web data or deliver it to Notion, budget those services separately. RohitAI’s guides to Firecrawl MCP free access and usage limits and Apify-to-Notion permissions and recurring costs address those adjacent choices; Zapier tasks do not establish either service’s full cost.

What should you check before the cutover?

Inventory Next Gen workflows, external SDK jobs and SDK-enabled Code steps. Record expected run volumes, repeated actions, AI tiers and tool-call counts. Give the billing owner a forecast for the whole account, with headroom for variable demand.

Plan for limited visibility: Next Gen has totals but no per-run task counts or Task History yet. External SDK usage appears as one account-level total, without a script-by-script breakdown. Zapier’s billing page is the account-wide reference; its UTC periods can differ from local-time Zap history.

Keep per-job operation estimates where useful, then reconcile them with billing totals after December 1. Those estimates are planning aids, not an invoice. If nonessential work can stop, agree when to pause it. If it cannot, settle the capacity and overage decision before it reaches the allowance.