See what changed.
Before your users do.
The Pixel House is visual regression testing and screenshot monitoring. API-first, with MCP for your AI coding assistant. From £19/month.
5,000 free screenshots per month. No credit card required.
How it works
Three steps to visual confidence
Capture
Paste a URL. Get pixel-perfect screenshots across desktop, tablet, and mobile viewports in seconds.
-d '{"url":"https://example.com"}'
Compare
We diff every pixel against your baseline using SSIM perceptual comparison. Not just pixels — structural similarity. Dramatically fewer false positives.
Threshold: 1.0% · SSIM: 0.998
Alert
Get notified the moment something breaks. Slack, email, or webhook. Schedule monitoring hourly, daily, or on every deploy.
/pricing · 3.2% diff · Layout shift
API-first
Built for your workflow, not ours
$ curl -X POST https://api.thepixelhouse.co.uk/v1/tests/run \
-H "Authorization: Bearer ph_live_xxx" \
-d '{
"url": "https://example.com",
"viewport": "desktop",
"baseline": "bl_abc123"
}'
{
"status": "pass",
"diff_percentage": 0.02,
"ssim_score": 0.998,
"diff_url": "https://cdn.thepixelhouse.co.uk/diffs/..."
}
A single POST request captures a screenshot. A second request compares it against your baseline. Works with any language, any framework. Integrate with your CI/CD pipeline, or use the MCP server from your AI coding assistant.
No browser dependencies to install. No flaky Selenium scripts. Just an API key and an HTTP client. We handle the rendering infrastructure so you can focus on catching regressions, not managing headless browsers.
Visual diffing
Pixel-level accuracy.
Perceptual intelligence.
SSIM-based comparison catches real regressions while ignoring sub-pixel font rendering noise. No more 3am false positive alerts.
Traditional pixel diffing flags every anti-aliasing change as a regression. SSIM (Structural Similarity Index) evaluates luminance, contrast, and structure the way the human visual system does, reducing false positives by an order of magnitude. Set your own threshold, define ignore regions, and approve changes directly from the dashboard.
MCP integration
Works inside your AI coding assistant
The Pixel House exposes a Model Context Protocol server so AI coding assistants can capture screenshots, run comparisons, and check baselines without leaving your editor. Ask your assistant to check for visual regressions after a refactor and it handles the rest.
Compatible with Claude Code · Cursor · Windsurf · GitHub Copilot
Pricing
Flat rate. No per-screenshot surprises.
Solo
- 10 projects
- 5 monitor schedules
- 30-day retention
- 1 user
- MCP server access
Team
- Unlimited projects
- 25 monitor schedules
- 90-day retention
- 5 users included
- MCP server access
- GitHub Action
Agency
- Unlimited projects
- Unlimited schedules
- 1-year retention
- 15 users included
- MCP server access
- Priority support
“No per-screenshot billing. Test more, not less.”
How we compare
Honest numbers
| Feature | Pixel House | Competitor A | Competitor B | Competitor C |
|---|---|---|---|---|
| 50K screenshots/mo | £49 | ~$400 | ~$299 | ~$969 |
| MCP server | Yes | partial | No | No |
| Scheduled monitoring | Yes | No | No | No |
| No CI/CD required | Yes | No | No | No |
| SSIM perceptual diffing | Yes | No | No | AI-based |
| Flat-rate pricing | Yes | No | No | No |
From the blog
Guides and benchmarks
- What Is an MCP Server? A Plain-Language Guide for DevelopersAn MCP server is a program that gives an AI assistant new tools. With 10,000+ public servers now live, here is how the Model Context Protocol works.
- SSIM vs Pixel Diff: A 40-Site Benchmark of Our EngineSSIM is said to eliminate the false positives that plague pixel diffing. We benchmarked both on 40 real sites with our own engine. The result surprised us.
- From Prompt to Verified UI: An MCP Workflow Case StudyHow we use our own MCP server to catch visual regressions in AI-generated code, walked through one real change to our pricing page that a prompt quietly broke.
See what changed.
5,000 free screenshots. No card required.