AI just changed your website.
Nobody told you.
Coding agents, developers, plugins and quick fixes ship edits to your site every day. Snapvisor photographs your pages, shows you exactly what moved, and nothing becomes the new normal until a person approves it — on a pull request, or on your live pages with no code at all.
14-day Pro trial · Works with your developer’s tools, or with none at all
The approval step your website never had
Fits the test suite you already run
The blind spot
Your site keeps changing. Who approved that?
Code changes, AI edits and updates alter what your visitors see. Snapvisor photographs the result, shows what moved, and makes a person say yes before it counts.
An AI agent changed it
Claude or Cursor shipped a fix overnight and the pricing button moved. Nobody asked for that, and nobody saw it.
A teammate's CSS leaked
A tweak two components away nudges your checkout button below the fold. Every test still passes.
An update repainted it
A dependency bump or a theme change turns your primary button off-brand. No assertion ever checks the hue.
Choose your path
Two ways in. One approval step.
Same engine, same approval step. Pick the door that fits how you work.
I just want to watch my site
- Tell us which pages
- We photograph them and compare with the last look you approved
- You see what moved and approve or flag it
No code. No install. Public pages, up to 10 per check.
I ship through a pipeline
- Add one step to the CI you already run
- Every pull request gets a visual status check
- Nothing merges until a person approves the changes
GitHub, GitLab, Playwright, Storybook, Cypress, Puppeteer.
Live regressions
What a regression actually looks like.
Three visual failure modes your team can catch before merging — even when logic tests pass.
Layout shift
A CTA drifts out of place after an unrelated CSS change. Pixels move; tests stay green.
Color regression
A token update repaints a primary button off-brand. No assertion ever checks the hue.
Text overflow
A longer string clips inside a fixed button. Logic is fine — the label is unreadable.
How it works
Visual regression testing, made plain: three steps, one review.
Whether your developer’s tools or Snapvisor itself takes the pictures, the loop is the same.
Capture
Your developer’s tools upload screenshots from Storybook, Playwright or Cypress, or Snapvisor photographs your public pages itself. No developer? Just name the pages.
Diff
Each screenshot is compared against an eligible baseline with a perceptual engine that reduces rendering noise.
Review
Changes wait for a person. Approve the intended ones, flag the rest. On a pull request, that decision is the status check.
Built for confidence
Everything you need to ship UI without fear.
From pixel-precise diffing to a review flow your whole team will actually use.
Pixel-perfect diffing
Perceptual comparison with anti-aliasing detection and adjustable thresholds for visual changes.
Cross-browser & responsive
Capture your configured breakpoints across Chromium, Firefox, and WebKit with Playwright.
Review workflow
Approve or reject changes inline. Approved builds can become baselines for later comparisons.
PR status checks
Post visual results to your pull request and require the check in your branch protection rules.
Parallel & fast
Shard screenshot uploads across CI jobs and bring the results together in one review.
Notifications
Slack and email notifications keep your team informed when visual changes need a review.
MCP for AI agents
A remote MCP server lets Claude Code, Cursor, and other agents read diffs and approve changes.
In your CI
One step in the pipeline you already run.
Snapvisor plugs into GitHub Actions, GitLab CI, CircleCI, and anything that runs your tests.
- Install one @snapvisor SDK — no infrastructure to run.
- Screenshots upload from the CI job you already have.
- A status check posts straight to the pull request.
- Baseline updates automatically when you merge to main.
# .github/workflows/visual.yml
- name: Upload screenshots to Snapvisor
run: npx @snapvisor/cli upload ./screenshots
env:
ARGOS_TOKEN: ${{ secrets.ARGOS_TOKEN }}
# Example: 412 screenshots uploaded
# ✓ 409 unchanged
# ! 3 changes need review →AI agents
Let AI agents review your UI.
Snapvisor runs a remote MCP server. Claude Code, Cursor, and any MCP client can create builds, upload screenshots, run visual diffs, and approve changes — no CI pipeline required.
$ claude mcp add --transport http snapvisor https://mcp.snapvisor.ioBuilds & projects
List and inspect builds, projects, and their status across your organization.
Screenshot diffs
Read the visual changes in any build — baseline, current, and the computed diff.
Reviews & approvals
Create reviews to approve or reject visual changes without leaving the agent.
Comments
Read and post build comments so humans and agents stay on the same thread.
OAuth or a personal access token, tools derived from the same public API. Read the connect guide Or set up your coding agent
For the engineers
Enforced, not advisory.
Agents can propose. People approve. This is what actually holds the line.
A required status check
Mark the snapvisor check as required in branch protection and an unapproved visual change cannot merge.
Approvals are recorded
Every approval names the person, and the agent that acted for them, in the build's review history.
Agents hold only the keys you give them
Agent access is scoped per connection and revocable in one click. Switch a project to “Require a person to approve” and an agent’s approval is refused.
Roles decide who may approve
Viewer, reviewer and admin levels per project. SSO and SAML for the team.
Smart thresholds
Tuned to notice what humans notice.
Snapvisor combines anti-aliasing detection with adjustable perceptual thresholds to reduce rendering noise. Tune comparisons around the changes your team needs to review. Drag this example threshold to see how sensitivity changes what gets flagged.
Full coverage
Your browsers. Your breakpoints.
Configure Playwright to capture your browser and viewport matrix, then review the screenshot diffs together.
Why Snapvisor
Turn screenshot tests into team reviews.
Moving off Lost Pixel, whose team announced its sunset? Read the Lost Pixel comparison. Already running screenshot assertions? See how Snapvisor fits alongside Playwright's built-in screenshot tests. Switching over? Follow the Percy migration guide, the Chromatic migration guide, the Applitools migration guide, the Playwright toHaveScreenshot migration guide or the Lost Pixel migration guide.
Integrations
Works with the tools you already have.
Connect your existing test suite and CI with screenshot SDKs and integrations.
Team control, open at the core.
Built on a proven open-source engine, with access control and single sign-on. Contact us to arrange self-hosting access, or read how self-hosted visual regression testing works.
Pricing
Start free. Scale when you do.
New customers start with a 14-day Pro trial. Pro is usage-based — you only pay for what you render beyond the included screenshots.
Free
For open source and side projects.
- 7,000 screenshots / month
- Unlimited projects
- Watch your public pages (up to 10 per check)
- GitHub & GitLab status checks
- Every MCP tool for AI agents
- Fine-grained access control
- Community support
Pro
Usage-based, with a 14-day free trial.
- 15,000 screenshots included, then usage-based
- Review workflow & approvals
- Slack & email notifications
- GitHub SSO included on Free & Pro
Enterprise
For org-wide scale and control.
- Volume screenshot pricing
- SSO / SAML
- Self-hosting by arrangement
- Role-based access control
- Dedicated support
Prices shown in USD.
FAQ
Questions, answered.
Playwright supports screenshot baselines and configurable thresholds. Snapvisor adds a shared review dashboard, perceptual diffs, and pull-request status checks, so your team can approve intended changes and track regressions together.
The diff engine uses anti-aliasing detection and perceptual thresholds to reduce rendering noise. Set a threshold when capturing a screenshot to tune how sensitive its comparison should be.
No. Point Snapvisor at your existing Storybook, Playwright, Cypress, or Puppeteer suite using the @snapvisor SDKs — open-source, drop-in forks of the official @argos-ci SDKs — and it captures screenshots from the stories and pages you already have.
When a change is detected, the pull request gets a status check linking to a side-by-side review. Approved builds become eligible baselines, and prior approvals help your team focus on new changes.
Yes. Snapvisor runs a remote MCP server at mcp.snapvisor.io. Any MCP client — Claude Code, Cursor, and others — can create builds, upload screenshots, run visual diffs, and approve or reject changes — no CI pipeline required. Every MCP tool is included on every plan, Free included: the build loop, reviewing and approving changes, comments and team tools.
Snapvisor builds on a proven open-source visual-diff engine and publishes its SDKs. The hosted service brings the review dashboard, integrations, and MCP server together. Contact us to arrange access to the Snapvisor server for self-hosting.
Yes. Add the public pages of your website under your project’s “Pages to watch”, or ask Claude to do it, and Snapvisor photographs them, compares them with the last version you approved and shows what moved. No code and nothing to install. Pages behind a login and private addresses are not supported.
Know what changed. Decide what ships.
Watch your website or guard your pull requests. Either way, a person has the last word.
14-day Pro trial · Cancel anytime