The best visual regression testing tools in 2026

Visual regression testing catches unintended changes to your UI by comparing screenshots (or richer snapshots) of the current build against a known-good baseline. We make Dungbeetle — so yes, it's on this list. The rest of the list is honest anyway; every tool here is the right choice for someone.

Three questions narrow the field fast. Do you test components (Storybook stories) or whole pages? Chromatic dominates the first; most others do the second. Do you want a hosted review dashboard or baselines in your repo with no server? And is a pixel diff enough, or do you need to know what changed structurally? Budget matters too: per-snapshot metering is the norm, and it's where bills surprise people.

Summary

Tool	Best for	Pricing headline	Hosting	Maintained?
Playwright toHaveScreenshot	Teams already on Playwright	Free (Apache-2.0)	Local / your CI	Yes (Microsoft)
Applitools	Enterprise cross-browser Visual AI	Sales-quoted; small free tier	Hosted SaaS	Yes
Argos	PR-centric review on a budget	Free 5k/mo; Pro from $100/mo	Hosted, open-source core	Yes
BackstopJS	Free local page-level pixel diffs	Free (MIT)	Local / your CI	Dormant — last release Sep 2024
Chromatic	Storybook component testing	Free 5k/mo; from $179/mo	Hosted SaaS	Yes (Storybook team)
Lost Pixel	Nobody new — sunsetting	n/a	Hosted + OSS mode	No — team joined Figma, Apr 2026
Percy	Cross-browser page screenshots in CI	Free 5k/mo; paid via BrowserStack	Hosted SaaS	Yes (BrowserStack)
Dungbeetle	Semantic diffs, agents, flat pricing	Free CLI; cloud free in closed beta	Local-first + optional cloud	Yes (that's us)

Playwright's built-in toHaveScreenshot

If you already write Playwright tests, await expect(page).toHaveScreenshot() is visual regression testing you get for free. It captures a screenshot, retries until the page stabilizes, and compares against a baseline stored next to your test file, with mask, maxDiffPixels, and threshold options to tame flake. It's maintained by Microsoft as part of Playwright itself, so it will never be abandoned while Playwright lives.

The genuine strength is zero new moving parts: no account, no upload, no per-snapshot bill, baselines reviewed in the same PR as the code. The limits show up at scale. Screenshots render differently across OSes and browsers, so you need per-platform baselines (most teams standardize on Linux CI). There's no review UI beyond the HTML report and no approval workflow — updating baselines is just a --update-snapshots commit.

Pick it if you're on Playwright, your suite is modest, and you'd rather own the workflow than buy one. Pricing: free, part of Playwright (docs, as of July 2026). See also our deeper Dungbeetle vs Playwright comparison.

Applitools

Applitools Eyes is the enterprise heavyweight. Its "Visual AI" comparison engine is genuinely differentiated: instead of raw pixel math, it analyzes layout, content, and color perceptually, which cuts false positives from anti-aliasing and rendering noise better than anything else on this list. The Ultrafast Grid renders your app across many real browsers and viewports from a single test run, and it supports web, mobile, PDFs, and more.

The trade-off is cost and process. There is no public rate card — pricing is sales-quoted around "Test Units," plans beyond a small free tier (around 100 checkpoints/month) require annual commitment, and third-party buyers report entry contracts near $200/month (platform pricing, as of July 2026). That's fine for enterprises with procurement; it's a poor fit for a five-person team that just wants diffs in PRs.

Pick it if you need the best false-positive suppression in the industry, true cross-browser coverage, and you have the budget and patience for a sales cycle.

Argos

Argos is a hosted visual testing platform with an open-source core (argos-ci/argos). It slots into CI, compares screenshots from any framework (first-class Playwright and Storybook support, plus ARIA snapshot comparison), and posts clear diffs as PR status checks. Its real strength is honest, simple economics for a hosted review workflow: a free Hobby tier with 5,000 screenshots/month, and Pro from $100/month including 35,000 screenshots with $0.004 overage — cheaper per-screenshot than Percy or Chromatic (pricing, as of July 2026).

Limits: it's still per-screenshot metering, the comparisons are pixel-based, and it's a smaller company than BrowserStack or Chromatic's — weigh vendor risk accordingly. The open-source core softens the lock-in concern since you can self-host.

Pick it if you want Percy-style PR review at a friendlier price and like supporting open-source infrastructure.

BackstopJS

BackstopJS is the veteran of free, local, page-level visual testing: config-driven scenarios, Puppeteer/Playwright rendering, pixel diffs, and a genuinely good self-contained HTML report with side-by-side and scrubber views. It's MIT-licensed, runs entirely on your machine or CI, and costs nothing forever. For a decade it was the default answer to "free visual regression testing."

The problem in 2026 is momentum. The last npm release was 6.3.25 in September 2024, and repository activity has been quiet since (github.com/garris/BackstopJS, as of July 2026). It still works — the tool is mature and the scope is stable — but you're adopting software with an uncertain maintenance future, an aging engine story, and no review workflow beyond the report.

Pick it if you want a free local pixel tool today, accept the maintenance risk, and your needs fit its page-scenario model. If you're already there and looking to move, we maintain a BackstopJS migration guide.

Chromatic

Chromatic is made by the team that maintains Storybook, and for component-level visual testing it is simply the strongest tool available. It snapshots every story in cloud browsers, localizes regressions to individual components, and its UI Review workflow (assign reviewers, discuss diffs, approve baselines) is a real collaboration product designers actually use. TurboSnap uses your git history to skip unchanged stories and cut costs.

The limits are the mirror image: it's built around Storybook (Playwright and Cypress E2E support exists but is not the center of gravity), it's component-first rather than whole-page-first, and per-snapshot metering across stories × viewports × browsers adds up. Pricing: free tier with 5,000 snapshots/month (Chrome only), Starter $179/month for 35,000 snapshots, Pro $399/month for 85,000, overage $0.008 per snapshot (pricing, as of July 2026).

Pick it if your team lives in Storybook — nothing else matches it in that lane. More detail in Dungbeetle vs Chromatic.

Lost Pixel

Lost Pixel was a well-liked open-source visual testing tool with a hosted cloud, covering Storybook, Ladle, and page screenshots with a pragmatic GitHub-native workflow. It deserved its following: simple config, fair pricing, responsive founders.

Don't adopt it in 2026. On April 22, 2026 the team announced they are joining Figma and sunsetting Lost Pixel (announcement), and the GitHub repository has been archived. The open-source mode still installs and runs, but there will be no fixes, no dependency updates, and the platform is winding down.

If you're a Lost Pixel user, your realistic options are the other tools on this page — we wrote a dedicated Lost Pixel migration guide that maps its config to Dungbeetle's, and Argos or Chromatic are natural homes for its Storybook users.

Percy (BrowserStack)

Percy, owned by BrowserStack, is the best-known hosted visual testing platform for page-level screenshots in CI. Its standout strength is true cross-browser rendering backed by BrowserStack's device cloud — the same snapshot rendered in real Chrome, Firefox, Safari, and Edge at multiple widths — plus a polished review dashboard and SDKs for nearly every test framework.

The costs come from that same model: every browser × width combination is a billed screenshot, so coverage multiplies your bill quickly, and reviews happen in Percy's dashboard rather than your repo. Pricing: a free tier with 5,000 screenshots/month with unlimited users and projects; paid plans are volume-based through BrowserStack, with entry plans historically around $199/month (plans and billing, as of July 2026).

Pick it if cross-browser rendering fidelity is a hard requirement and you want an established vendor. See Dungbeetle vs Percy.

Dungbeetle

Dungbeetle is our tool, so calibrate accordingly. It's a snapshot and visual regression testing CLI — free forever, source-available under FSL-1.1 (not open source, and we won't pretend otherwise) — that pairs semantic JSON diffs with pixel-compared screenshots. Instead of only telling you pixels changed, it diffs structured state: DOM, terminal output, API responses, desktop apps, and game state, alongside screenshots with tolerance. Baselines commit to your repo, so like Playwright's built-in there's no server required; unlike it, the diff tells you what changed.

The other bet is agents. Dungbeetle is agent-native: device-flow auth, scoped tokens, an MCP server for Claude Code, Cursor, Codex, and Gemini CLI, and human-gated baseline promotion so an AI agent can run and read visual tests but can't silently bless a regression. The optional managed cloud is in closed beta (free during beta) with proposed flat tiers of $0/$29/$99/$249 — unlimited seats, no per-snapshot charges, storage-only overage at $0.10/GB-month.

Honest limits: the cloud isn't generally available yet, the ecosystem is younger than everything above, and if you need real cross-browser rendering fidelity, Percy or Applitools do that better today. Try it without signing up: npx dungbeetle anon <url> --compare <url>. Migration guides for Lost Pixel, Wraith, BackstopJS, and Loki live at /migrate/.

How to choose

• You live in Storybook → Chromatic. It's the native path, built by the Storybook maintainers.
• You need true cross-browser rendering → Percy for CI screenshots at scale, Applitools if you also want the best AI-powered noise suppression and have enterprise budget.
• You're on Playwright and want free → toHaveScreenshot. Accept per-platform baselines and no review workflow.
• You want hosted PR review without enterprise pricing → Argos.
• You want semantic diffs, non-web surfaces (terminal, API, desktop, games), agent workflows, or flat pricing → Dungbeetle.
• You want a free local pixel tool and accept maintenance risk → BackstopJS.
• You're on Lost Pixel → migrate; the product is sunsetting. Start at /migrate/.

Last verified: July 2026. Deeper head-to-heads: Dungbeetle vs Percy · Dungbeetle vs Chromatic · Dungbeetle vs Playwright · Migration guides.