Applitools pricing in 2026: the checkpoint cost model.

What a checkpoint is

Per the Applitools documentation (applitools.com/pricing), a checkpoint is consumed every time the Eyes SDK runs a visual verification at one specific configuration. A test that calls eyes.check() once on a single browser at a single viewport consumes one checkpoint. A test that calls the same method across three browsers (Chromium, Firefox, WebKit) at two viewports (desktop, mobile) consumes six checkpoints.

The economics therefore look multiplicative. A team running 100 tests, each with five checkpoint calls, across three browsers and two viewports, consumes 100 × 5 × 3 × 2 = 3,000 checkpoints per full run. At one full run per pull request and 1,000 PRs per month, the monthly consumption is 3 million checkpoints. That is a meaningful enterprise tier; the same coverage at one browser and one viewport drops to 500,000 checkpoints.

The unit is also one of the things that makes the Applitools versus other visual-regression vendor comparison hard to do casually. Percy prices by snapshot (a simpler unit); Chromatic prices by snapshot taken in Storybook. Comparing on list price alone without normalising for the unit definition produces misleading conclusions.

Published tiers

The vendor pricing page (applitools.com/pricing) publishes tiers including a free trial, Starter, Eyes, and Enterprise levels. The published tiers scale with checkpoint volume and feature scope (concurrency, integration depth, support level). Enterprise tier handles custom checkpoint volumes, dedicated support, and the additional security and deployment options.

As of April 2026 the exact per-month rates for the published tiers are on the vendor page; buyers should reference the live page rather than a third-party summary because the rates and tier feature splits change. The structural shape (tier-bands by checkpoint volume) is stable.

What affects the total beyond list price

Concurrency. Running visual regression on every pull request requires concurrent test execution; the parallel-runner capacity affects the tier. Enterprise contracts typically include negotiated concurrency.

Multi-region testing. Some customers test against region-specific deployments (EU, US, APAC) and consume checkpoints separately for each. This multiplies the consumption.

Branch coverage. Teams that run visual regression on long-lived feature branches consume baseline-creation checkpoints separately from main-branch comparisons. The economics differ from a single-main-branch workflow.

Component-level versus page-level checks. A team that uses eyes.checkRegion() on individual components inside a page consumes more checkpoints than one that takes a single full-page screenshot. Component-level is more precise but more expensive in checkpoint terms.

Hidden costs

Reviewer time. Every flagged diff is human time to accept or reject. Applitools' Visual AI reduces false positives but does not eliminate them. Budget reviewer time as part of total cost; for a team running 500 PRs a month with 10 percent flag rate, that is 50 review decisions per month per reviewer.

Baseline drift. Baselines must be maintained or they accumulate intentional changes that mask future regressions. Teams should have a baseline-review cadence (often quarterly) that consumes engineering time.

SDK integration. The first integration into an existing test suite is straightforward (the SDK is well-documented) but tuning it for the team's false-positive tolerance and configuration density takes real engineering iteration.

How to budget before the sales call

The honest method: estimate annual checkpoint consumption first, then apply the published tier rate band, then add 20 percent for the configuration count creeping during adoption. The composite is a defensible budget figure that lets the team start the procurement conversation with a number rather than a question.

For a typical small team (100 PRs per month, 50 tests, 5 checkpoints each, 1 browser, 1 viewport): 100 × 50 × 5 × 1 × 1 = 25,000 checkpoints per month. Whether that falls in Starter, Eyes, or Enterprise tier depends on the published thresholds for the current quarter; check the live vendor page.

For a typical enterprise team (1,000 PRs per month, 500 tests, 10 checkpoints each, 3 browsers, 2 viewports): 1,000 × 500 × 10 × 3 × 2 = 30 million checkpoints per month. This is firmly enterprise tier with custom negotiation; the per-checkpoint rate at this scale is materially below the published Starter rate.

When Applitools is worth the unit cost

Applitools is worth the unit premium when the team's false-positive load on visual regression is meaningful and the Visual AI classification demonstrably reduces it. The honest test is a one-week pilot under the team's real test suite, with the reviewer time logged carefully. If Applitools cuts the review hours by 30 percent or more, the unit premium pays for itself; if it does not, a snapshot-based competitor may be more cost-effective at equivalent coverage.

See Applitools vs Percy for the head-to-head with the most common alternative and AI visual regression overview for the broader category landscape.

Frequently asked questions

What counts as one Applitools checkpoint?

A checkpoint is one visual verification at one configuration (browser, viewport, region). A single test that takes one screenshot across three browsers and two viewports consumes six checkpoints. The checkpoint definition is specified in the vendor documentation and matters because the bill scales with it.

Is Visual AI just hype?

Visual AI is the vendor's name for an analysis layer above pixel-diffing that classifies changes (content, layout, style, movement) and ignores cosmetic noise like one-pixel anti-aliasing. It is a real feature that reduces false positives compared with pure pixel diffing. Whether it justifies the per-checkpoint premium depends on the customer's false-positive volume and reviewer time.

Can I run Applitools self-hosted?

Enterprise tiers include self-hosted and dedicated cloud deployments per the vendor. The published default is cloud SaaS. Buyers in regulated industries should request the current architecture options as part of the procurement conversation.

How does this compare to Percy pricing?

Percy prices by snapshot, with a simpler unit definition. For the same coverage profile, Percy and Applitools can produce different effective per-month bills; the right approach is to normalise on coverage (what flows, what configurations, what review effort) rather than on raw unit count. See the Applitools vs Percy comparison for the structural framing.

Are there volume discounts?

Yes, per the vendor's published enterprise tiers. The discount structure is custom and tied to checkpoint volume bands. Buyers approaching tier boundaries can sometimes negotiate the better rate before crossing the threshold.