iiinova — where your digital self keeps evolving

You've approved the same diff 40 times. Argos noticed. When a visual change keeps reappearing and keeps getting approved, the bot starts ignoring it for you. Configurable threshold, on by default, every plan. Anything new or different still surfaces.

000

Custom domains are included in paid plans. Add one from Settings → Deployments → Domains. argos-ci.com/changelog/c...

000

The domain always serves your latest production deployment. Previews keep their own per-deployment URLs. And once it's live, it's the link GitHub shows: the commit status and the PR comment point at your domain.

100

The whole setup is one DNS record. No TXT record to add, nothing to verify by hand. Once it resolves, Argos issues and installs the SSL certificate automatically. Nothing to buy, upload, or renew.

100

Argos @argos-ci.com · 31/08/2026

Custom domains for deployments are here. Serve your production deployments from a domain you own. storybook.acme.com instead of a generated URL nobody recognises.

100

You don't need to hand-roll screenshot stabilization. Argos waits for fonts, images and aria-busy before every capture — then hides carets and scrollbars, pauses GIFs, disables animations and fixes antialiasing. The flake teams fight manually is already handled.

000

Argos @argos-ci.com · 30/08/2026

Most teams assert on UI like this: expect(button).toHaveClass('primary') That passes when the button is invisible, off-screen, or the wrong color. A screenshot asserts what the user actually sees — all of it, in one line. The assertion you skip is the bug you ship.

000

Argos @argos-ci.com · 29/08/2026

We put a 99.99% uptime SLA on Enterprise because a visual testing service that's down blocks every merge in your org. CI tooling has one job: be there when the build runs.

000

Argos @argos-ci.com · 28/08/2026

Every team screenshots the happy path. Nobody screenshots: ◆ empty state ◆ loading skeleton ◆ error state ◆ a 200-character username ◆ zero results ◆ offline That's where the layout actually breaks.

000

Argos @argos-ci.com · 27/08/2026

On a financial chart, the data is the pixels. A one-pixel shift can be a real bug. Vojtech Gintner, CTO of Finviz: "We use Playwright for E2E, and Argos adds the missing layer of visual testing. On our charts, a single pixel difference can indicate a real bug."

000

Argos @argos-ci.com · 26/08/2026

One command gives your coding agent a visual review workflow: npx skills add argos-ci.com Then just ask: "Use $argos-pr-review to review this pull request with its Argos build."

000

Argos @argos-ci.com · 25/08/2026

Playwright locators, best to worst: getByRole('button', { name: 'Save' }) getByLabel('Email') getByText('Welcome back') getByTestId('save-btn') locator('.btn > span:nth-child(2)') If a locator breaks when you restyle, it was testing the wrong thing. argos-ci.com/blog/playwr...

000

Argos @argos-ci.com · 24/08/2026

An agent can point at the exact pixel. argos comment create 1234 \ --diff <diffId> \ --anchor-point 412,168 \ --body "Focus ring disappeared here." Or --anchor-lines 40,48 to pin a range inside a JSON or Markdown snapshot.

Argos @argos-ci.com · 23/08/2026

"Pixel-perfect diffing" sounds rigorous. In practice one global threshold means you either drown in noise or miss real bugs. A marketing hero and a financial chart don't deserve the same tolerance. Sensitivity belongs per screenshot.

000

Argos @argos-ci.com · 22/08/2026

Don't test 12 viewports. Test 3 that mean something: 375px — the phone where things break 768px — the tablet nobody QAs 1440px — what you designed on More viewports isn't more coverage. It's more diffs to review.

000

Argos @argos-ci.com · 21/08/2026

How does your team catch UI regressions today? ◆ Automated visual tests ◆ Manual QA before release ◆ Preview deploys and eyeballs ◆ Users tell us Genuinely curious where most teams land.

000

Argos @argos-ci.com · 20/08/2026

Delete ARGOS_TOKEN from your repo secrets. Argos authenticates uploads with GitHub OIDC — short-lived, GitHub-signed identity tokens. Add id-token: write and drop the secret. Forked PRs fall back to a verified tokenless flow.

000

Argos @argos-ci.com · 19/08/2026

We rebuilt Argos's UI components on React Aria. Accessibility, keyboard behavior and focus management we no longer hand-roll — and a large amount of custom code deleted. What we learned: argos-ci.com/blog/react-...

000

Argos @argos-ci.com · 18/08/2026

If every test logs in through the UI, you're paying for the same login 200 times. Log in once, save storageState, reuse it everywhere. In most slow Playwright suites this is the single biggest win available. argos-ci.com/blog/speed-...

000

Argos @argos-ci.com · 17/08/2026

We're SOC 2 compliant. The honest summary for a small team: most of the work isn't adding security. It's writing down the security you already had, then proving it every quarter. argos-ci.com/blog/soc-2

001

Argos @argos-ci.com · 16/08/2026

Le Monde's CMS, Sirius, is the backbone of the group's content operations. A delay in publication can cost millions of euros. Paul Laleu, CTO: "Thanks to Argos, we can confidently deploy updates to Sirius without worrying about unexpected issues."

000

Argos @argos-ci.com · 15/08/2026

A screenshot tells you it looks right. An ARIA snapshot tells you it is right. - heading "Dashboard" [level=1] - button "New project" - list: - listitem: "Project A" Catches the div-soup regression no pixel diff ever will.

000

Argos @argos-ci.com · 14/08/2026

Visual review, from the terminal: argos build get 1234 argos build snapshots 1234 --needs-review argos review create 1234 --event approve Anything a human can do in the UI, an agent can do too. Same permissions, same audit trail.

000

Argos @argos-ci.com · 13/08/2026

Committing screenshots to your repo means: ◆ repo size that only grows ◆ binary diffs nobody can review ◆ "update snapshots" commits nobody audits ◆ baselines that only match one machine's fonts Baselines are build artifacts. Store them like build artifacts.

000

How it looks in practice: the agent opens the PR and the demo recording is already in the comments. A reviewer pins a spot, the agent reads the coordinates from the CLI, fixes, resolves. Following along on Product Hunt today: www.producthunt.com/posts/media...

000

Visual testing isn't only for pixels. Your API response shape, generated Markdown, and rendered email HTML all regress silently too. argos upload ./out --files '**/*.json' Same baseline workflow. No image involved.

000

Media sharing · Changelog

An image costs 1 screenshot unit, a video 25 — on the meter you already have. Live on every plan. We're on Product Hunt today, questions welcome: www.producthunt.com/posts/media... argos-ci.com/changelog/2...

argos-ci.com

Upload standalone images and videos from the CLI, SDK, API, or MCP, get a share link with ready-to-paste Markdown, and let Argos post them on the pull request automatically.

000

Reviewers pin comments to a point on the image. Your agent reads the coordinates back with `argos media comment list`, fixes, replies, and resolves — feedback about pixels, as data a terminal can act on.

100

Re-upload after review and the link doesn't change: a media is an identity, every upload a version. The Markdown already pasted in the PR shows the newest image — and the version a reviewer commented on survives underneath.

100

Upload with --branch while the work happens. The media is staged: real, shareable, listed. The moment a pull request opens for that branch, Argos publishes everything staged there and posts one managed comment.

100

Your coding agent can't drag a screenshot into a pull request — GitHub has no API for it. Now it doesn't need one: `argos media upload` turns any image or video into a share link, and Argos posts it on the PR by itself. We're live on Product Hunt today 🚀

100

The MCP server derives its tools from the API, so it went from 37 tools to 71 the moment these shipped. No client update. Your agent can now onboard a teammate or tune a project's flaky-change settings. argos-ci.com/changelog/m...

The CLI grew the matching commands: argos test list --project acme/web argos project update --project acme/web --summary-check auto argos account member list --account acme

100

Projects: read and update every setting, transfer a project, grant contributors access, list deployments, set the deployment domain. Automation rules too — list, create, replace, deactivate.

100

Teams: list members and change their role, send and cancel invites, rotate the invite link, open a team to a verified email domain.

100

Argos @argos-ci.com · 09/08/2026

The Argos API could read your builds and review them. Everything else — who's on the team, how a project is configured, which automations fire — lived only in the app. Thirty-four new endpoints close that gap.

100

MUI runs 2.5M+ screenshots per month through Argos. Five projects. 1,100+ screenshots compared on every build in the Material UI repo alone. Base UI is in there too — the library that powers shadcn/ui. argos-ci.com/customers/mui

Argos @argos-ci.com · 08/08/2026

AI writes UI faster than humans can review it. An agent can open six PRs before lunch. Every one touches the layout. Nobody is opening six preview deploys. The bottleneck moved from writing code to verifying what it looks like.

000

Argos @argos-ci.com · 07/08/2026

Your agent doesn't have to just approve or reject. It can hold a conversation. argos comment create <build> --body "..." argos comment react <build> <id> eyes argos comment resolve <build> <id> Reply, react, resolve, subscribe. A reviewer, not a webhook.

Argos @argos-ci.com · 07/08/2026

An agent approving a build: argos review create 1234 --event approve --body "Intentional spacing change, matches the PR." --event takes approve, reject or comment. Same permissions as a human. Same audit trail.