Argos Review (2026): Pricing, Features & Honest Verdict

Argos is visual regression testing built for teams whose code is now partly written by AI agents. It catches every pixel, Markdown and JSON change on a pull request and lets people and agents review the diff together. Best for: frontend teams running Playwright or Storybook in CI. Price: free to 5,000 screenshots, Pro from $100/mo. Rating: 8.5/10.

What is Argos?

Argos is a visual testing service that plugs into your CI pipeline and tells you exactly what a pull request changed on screen. You point it at Playwright, Storybook, Cypress or a static build, it captures screenshots on every run, and it compares them against the baseline from your main branch. When something moves, you get a side-by-side diff with the changed region highlighted, and you approve or reject it before the code merges.

The company launched in 2023 and is run by Greg Bergé and Jeremy Sfez. What makes this launch worth attention is the repositioning. Argos used to sell itself on pixels. It now sells itself on a problem that did not exist three years ago: agents generate more changes than humans can review. The homepage says so directly. The newest feature, Media Sharing, lets an AI agent drop screenshots and screen recordings straight onto a GitHub pull request as shareable links with ready-to-paste Markdown, so the agent shows you what it built instead of describing it. That is a small feature with a clear thesis behind it, and the thesis is correct.

What are Argos’s key features?

Change detection beyond screenshots

Most visual testing tools compare images. Argos compares images plus Markdown, JSON, HTML, CSS, XML and YAML. This matters more than it sounds. If your agent regenerates a config file, rewrites a README section, or changes an API response fixture, that shows up as a reviewable diff in the same interface as a button moving four pixels. You stop having two review surfaces, one for what the code looks like and one for what the code emits. For teams where agents touch generated files constantly, this is the feature that earns the subscription.

The stabilization engine

Flaky visual tests are the reason most teams abandon visual regression testing within a quarter. Anti-aliasing, font loading, animation timing and scroll position produce diffs that mean nothing, and after the twentieth false positive people start clicking approve without looking. Argos runs deterministic pixel diffing with a stabilization layer on top. The marketing example on the homepage shows the same test run producing 34 diffs without stabilization and 2 with it. We cannot independently verify that ratio, but the problem it targets is the correct one, and the fact that the pitch leads with noise reduction rather than detection sensitivity suggests the team has actually run this in anger.

Collaborative review with pinned comments

Reviews happen on the diff itself. You click a pixel and start a thread there, so “the price tag is hidden on mobile” is anchored to the price tag rather than buried in a PR comment thread. You can pull in a designer or a feature owner who has no business reading the code but every business looking at the screen. And the review thread accepts agent participation: a comment can be addressed to the agent that opened the PR, which then picks up the requested change. This is the part of the product that reflects how mixed human and agent teams actually work, and it is ahead of where most competitors are.

Media sharing for agent output

The August launch. An agent running in CI can convert any image or screen recording into a hosted link with Markdown ready to paste into a pull request. Free accounts get public links with 30-day retention. Pro gets team-scoped links with a year of retention, which is the version you want if the recordings show anything about unreleased work. It is a modest feature, but it closes a real gap: before this, an agent that built a UI change had no way to show you the result without a human running the branch locally.

Agent-ready CLI and REST API on the free plan

Worth calling out because plenty of vendors would have gated it. The CLI and the REST API are on the free Hobby plan, not held back for paying customers. So is GitHub and GitLab integration, and so is flaky detection. The free tier is a working product rather than a demo, which is the single best signal about how a company intends to treat you later.

How much does Argos cost?

Argos starts free and Pro starts at $100 per month. The Hobby plan is $0 forever and covers up to 5,000 screenshots a month with visual and snapshot testing, Storybook support, the CLI and REST API, flaky detection, GitHub and GitLab integration, and public media links with 30-day retention. Pro starts at $100 a month, includes 35,000 screenshots, and adds team collaboration, private deployment protection, Slack and Microsoft Teams notifications, year-long team-scoped media retention and Pro support. Enterprise is quote-based and adds SAML SSO, fine-grained access control, a 99.99% uptime SLA and a dedicated contact.

Pricing is usage-based rather than per-seat, which is the right model here and the company says so explicitly on the pricing page. You are not punished for letting a designer or a product manager look at a diff. That alone separates it from most of the category.

The arithmetic on the Pro plan has a wrinkle worth knowing. $100 for 35,000 screenshots works out to about $0.00286 per screenshot. Extra screenshots beyond the allowance cost $0.004, which is roughly 40% more per screenshot than the ones bundled into the plan. That is the normal direction for overage pricing, so it is not a scandal, but it means a team running consistently over 35,000 is better off talking to sales than drifting on overages. Storybook screenshots run the other way: overages there are $0.0015, which is cheaper per screenshot than the blended included rate, so a Storybook-heavy team is quietly getting a better deal than the headline number suggests.

One oddity on the pricing calculator: set it to zero screenshots and it still reports $100 a month, then advises you to contact sales for a custom plan. It is a minor interface bug rather than a pricing trap, but it makes the tool useless at the low end of the Pro range, which is exactly where a team deciding whether to upgrade would use it.

What does Argos charge for SSO?

This is the part of the pricing page we like least. GitHub SSO is a $50 per month add-on. SAML SSO is $200 per month. Both sit on top of the $100 Pro base, so a team that wants SAML is paying $300 a month before it captures a single screenshot beyond the allowance.

Charging separately for single sign-on is a common practice and a bad one. SSO is a security control, and pricing security as a premium feature pushes smaller teams toward shared logins. To be fair to Argos, SAML is included at the Enterprise tier, and $50 for GitHub SSO is at the cheap end of what the industry charges for this. But it is still a line item that should be part of the product. If you are evaluating Argos for a company with a security review process, budget $300 a month, not $100.

How does Argos compare with Chromatic on price?

Chromatic is the reference point in this category, and the comparison is favourable to Argos on every number we could check. Chromatic’s paid tiers start at $179 a month for Starter and $399 for Pro, against Argos at $100. On overages, Chromatic charges $0.008 per extra billed snapshot. Argos charges $0.004, exactly half.

Both have a free tier, both integrate with Storybook, and Chromatic has the deeper Storybook story given it is built by the Storybook maintainers. If your entire workflow is Storybook and nothing else, that relationship is worth something. But if you are running Playwright end-to-end tests, or you want the non-image file types diffed, or you simply want to pay half as much for overage, Argos wins the comparison on substance rather than on marketing.

Who is Argos best for?

Use Argos if you have a frontend with real visual surface area, you already run Playwright or Storybook in CI, and you have started letting coding agents open pull requests. That last condition is what tips it from useful to necessary. When a human writes the diff, the human knows what should have changed. When an agent writes the diff, nobody does until something renders. Also use it if you maintain an open-source project: 5,000 free screenshots a month covers a lot of ground, and the company runs a sponsorship programme for open-source work.

Skip Argos if your product is mostly backend, or if your frontend is a small internal admin panel where a visual regression costs nothing. Skip it too if you need SAML and cannot get budget past $300 a month, because that is your real entry price rather than the advertised $100. And if your team has never managed to keep a visual test suite green, fix the flakiness culture before buying a tool that will surface more of it.

Best Argos alternatives

Chromatic is the most direct competitor, built by the team behind Storybook. Free tier, then $179 a month for Starter and $399 for Pro, with extra snapshots at $0.008. Better if Storybook is the centre of your workflow, worse on price at every tier.

Percy from BrowserStack is the enterprise-adjacent option, with broad cross-browser coverage and the BrowserStack device cloud behind it. It publishes no pricing on its main pricing page, which tells you which end of the market it is aimed at. Choose it if you already buy BrowserStack and want visual testing folded into that contract.

Applitools sells AI-assisted visual validation with a free trial plus starter and enterprise plans. It is the most feature-heavy of the three and the most likely to require a sales call. Worth a look if you need visual testing across a large matrix of browsers and devices and have a budget to match.

Final verdict: is Argos worth it?

Yes, and it is one of the better-priced developer tools we have looked at this month. The free tier is a real product rather than a lead magnet, the paid entry point is $79 a month cheaper than the obvious competitor, overage costs half what Chromatic charges, and the pricing page publishes explicit per-screenshot rates you can verify with a calculator instead of a sales call. The product itself is mature, has been shipping since 2023, and has named customers using it in production.

The reservations are narrow. Charging $50 to $200 a month for single sign-on is the wrong call and effectively triples the entry price for any company with a security review. The overage rate is 40% above the included rate, so heavy users should negotiate rather than drift. And the agent-collaboration story, while directionally right, is newer than the rest of the product and will need another release or two before it is the reason you buy rather than a bonus on top of solid visual testing. Start on the free plan, run it against your real PR volume for a month, and check what the screenshot count actually looks like before committing to Pro.