Argos, often searched as Argos CI, is an open-source visual testing platform. Your own tests take the screenshots, Argos compares them with a baseline build chosen from your Git history, and your team approves or rejects the differences on the pull request. Design Bug Bot, from 21st, is a GitHub App that reviews the UI code in a pull request against the components, tokens, spacing and type scale your repository already uses, and suggests fix code.
Put simply, Argos reviews the screenshots your tests capture, and Design Bug Bot reviews the code that produced them. Use Argos when you want pixel-level protection for the pages and stories your tests already visit; add Design Bug Bot when the recurring problem is new UI that works but drifts from your design system.
Disclosure: This article is published by 21st, which makes Design Bug Bot. It draws on Argos and 21st documentation and pricing pages checked on September 14, 2026. It is not a head-to-head benchmark, and neither tool was scored.
Argos and Design Bug Bot at a glance
| Argos | Design Bug Bot | |
|---|---|---|
| Review job | Visual regression review of screenshots against a baseline build. Argos overview | Design review of changed UI code against your existing patterns |
| Input | Screenshots from Playwright, Cypress, Vitest, Storybook, WebdriverIO or Puppeteer, or any pipeline through the CLI. Argos overview | The pull request diff and the repository it belongs to |
| Where the UI is rendered | In your CI; Storybook stories are captured with Playwright. Storybook SDK | In a disposable sandbox, when an isolated render succeeds |
| What counts as correct | A baseline build picked from commit history. Baseline build | Your repository's components, tokens, spacing and type scale |
| AI | Bring your own agent through an MCP server, CLI, agent skills or REST API. Argos for agents | Built-in review; findings include suggested fix code |
| Merge gating | A commit status you can require in branch protection; one rejection blocks a build. Argos | Advisory: one updated comment and a GitHub check per reviewed commit |
| Git providers | GitHub and GitLab. GitLab integration | GitHub only |
| Source and pricing | MIT-licensed open source; hosted plans billed per screenshot. Argos pricing | 5 free successful reviews, then AI credits. Details |
What Argos does well
Argos is open in a literal sense. According to its overview, the whole platform, including the backend, frontend and screenshot comparison engine, is developed in the MIT-licensed argos-ci/argos repository.
It meets your tests where they are. SDKs cover Playwright, Vitest, Storybook, Cypress, WebdriverIO and Puppeteer, and the CLI accepts screenshots from any other pipeline. The Storybook SDK captures stories in your CI with Playwright and supports Storybook 8 through 11. Argos can also compare non-image files, flag flaky tests, and deploy a preview of your Storybook or static build on every pull request.
Baselines stay out of your repository. Argos picks a baseline build by analyzing your commit history, and its Playwright migration guide is written for teams that want to stop committing PNG baselines to Git.
Review is a team process. On the Argos site, you can request reviewers and track each verdict; one rejection blocks a build, otherwise one approval passes, whether a person or an agent submitted it. The result lands as a commit status you can require and a single pull request comment kept current.
Argos is also built for agents. It runs an official remote MCP server that can list builds, inspect screenshot diffs, approve or reject builds and post comments, limited by the scopes you grant. Its AI review guide pairs the CLI with an argos-pr-review agent skill: the agent finds the build, summarizes what changed visually, compares the screenshots with the pull request's intent, and suggests approving or requesting changes. Creating the review itself requires a user token.
Pixels matched the intent, but the code did not
Argos's agent workflow asks a sensible question: do these screenshots match what the pull request set out to do? That is a question about images and a description.
Picture a pull request titled "Add export button to reports." The screenshot shows a new export button on the reports page, so it matches the intent, and an agent or a person can approve it. The problem is in the source: the button is a fresh implementation with its own padding and font size, sitting next to the Button component the repository already has.
Design Bug Bot reviews that source. It checks the changed UI against your existing components, tokens, spacing and type scale, and looks for layout risks such as horizontal overflow, fixed widths and actions hidden on mobile, plus unnecessary dependencies. A finding explains the problem, points to the file and line when available, and includes suggested fix code you can inspect, copy or hand to a coding agent.
Coverage does not depend on a test visiting the page. Changed pages and components are picked from the diff and rendered in a React harness in a disposable sandbox. When the render succeeds, the review can include screenshots and visual checks at desktop and mobile widths; when it cannot, you still get code-backed recommendations without screenshots.
What Design Bug Bot does not do is Argos's job. It keeps no approved baselines, runs no cross-browser pixel regression and does not block merges. It is GitHub only and focused on React codebases.
Where each one fits
An agent approves a build that matches the description. Keep Argos and its agent review for the visual change. Add Design Bug Bot if approved builds keep introducing one-off components or off-scale spacing.
A new route that no test visits yet. Until a Playwright or Cypress test navigates to it, Argos has no screenshot of it. Design Bug Bot picks the page from the diff and reviews it in the same pull request; add the test afterward so Argos protects it from then on.
A GitLab monorepo or a non-React front end. Argos supports GitLab and works with whatever your test runner can screenshot. Design Bug Bot is GitHub only and React-focused, so Argos alone is the right call there.
A default branch that drifted between releases. Argos reviews each build it receives. Design Bug Bot can also review recent changes on the default branch manually, after a chosen number of commits, or on a schedule, and it can push a branch that opens a fix pull request.
Argos pricing
As listed on Argos pricing on September 14, 2026: the Hobby plan is $0 with up to 5,000 screenshots a month; Pro starts at $100 a month, includes 35,000 screenshots, and charges $0.004 per extra screenshot, or $0.0015 for Storybook screenshots; Enterprise is custom. Qualifying open-source projects can apply for sponsored access.
Design Bug Bot is not billed per screenshot. After 5 free successful reviews, reviews use AI credits from a subscription that includes AI; see the trial and billing details.
A side-by-side trial
- Choose a pull request that adds a component with a close existing equivalent, and make sure one of your Argos tests captures a page that uses it. Write down what each tool should catch.
- Run Argos as usual. If you use the argos-pr-review skill, save the agent's summary and its suggested decision.
- Enable Design Bug Bot on the repository and let it review the same pull request. Note which findings came with screenshots and which are code-backed.
- Sort every issue into three groups: found by both, found only as a pixel difference, found only in the source. The last two groups tell you whether you need one tool or both.
If you are coming from Lost Pixel, the Lost Pixel alternatives guide covers the move. For Storybook-first teams, compare Chromatic vs Design Bug Bot.
Frequently asked
- Is Argos CI open source?
- Yes. Argos documents that its whole platform, including the backend, frontend and screenshot comparison engine, is developed in the MIT-licensed argos-ci/argos repository, and it also sells hosted plans with a free Hobby tier. Design Bug Bot is a GitHub App run by 21st, with each review in a disposable sandbox.
- Can Argos review visual changes with AI?
- Argos supports bring-your-own agents through an official MCP server, its CLI, installable agent skills such as argos-pr-review, and a REST API. An agent can summarize a build's visual changes, compare them with the pull request's intent and, with a user token, create an approve or request-changes review. Design Bug Bot is itself the reviewer and posts findings with suggested fix code.
- Does Design Bug Bot need the screenshot tests Argos relies on?
- No. Argos compares screenshots that your tests or pipeline upload. Design Bug Bot picks changed pages and components from the diff and renders them in a React harness; when that works, the review can include desktop and mobile screenshots, and when it does not, you get code-backed findings. It keeps no baselines, so it does not replace Argos regression checks.
- How much does Argos cost?
- As listed on Argos pricing on September 14, 2026, Hobby is free with up to 5,000 screenshots a month, Pro starts at $100 a month with 35,000 screenshots and $0.004 per extra screenshot ($0.0015 for Storybook screenshots), and Enterprise is custom. Qualifying open-source projects can apply for sponsored access.