Visual testing tools all take a screenshot and diff it. Only Frontguard explains why something changed, verifies a fix, and stays open source and self-hostable.
| Capability | Frontguard | Percy | Chromatic | BackstopJS | Lost Pixel | Argos |
|---|---|---|---|---|---|---|
| Open source | ✓ MIT | not supported | partial support | supported | partial support | ✓ MIT |
| CLI-first | supported | not supported | not supported | supported | supported | supported |
| AI change classification | supported | partial support | not supported | not supported | not supported | not supported |
| AI fix verification | supported | not supported | not supported | not supported | not supported | not supported |
| Anti-flake rendering | supported | partial support | partial support | not supported | not supported | partial support |
| Cross-OS render normalisation | supported | supported | supported | not supported | not supported | not supported |
| Self-hostable | supported | not supported | not supported | supported | partial support | partial support |
| Storybook integration | supported | supported | supported | not supported | supported | supported |
| MCP server for in-IDE agents | supported | not supported | partial support | not supported | not supported | not supported |
| PR comment with thumbnail triplet | supported | supported | supported | not supported | partial support | supported |
| Enterprise SSO/SAML | partial support | supported | supported | not supported | not supported | supported |
| Free tier | Forever | 5k/mo | 5k/mo | Free | not supported | 5k/mo |
| Pro entry | $29/mo | $199/mo | $179/mo | n/a | n/a | $100/mo |
| Snapshot overage | Spend cap | Quote | $0.008 | n/a | n/a | $0.004 |
| Actively maintained | supported | supported | supported | ✕ quiet | not supported | supported |
Every cell traces to documented vendor behaviour — see docs/research.md.
The honest version — what each tool is genuinely good at, and where Frontguard pulls ahead.
Polished hosted dashboard, broad framework SDKs, and mature review workflows backed by BrowserStack.
CLI-first and free forever — no per-screenshot billing that punishes a growing suite. Plus AI explanations, not just a red diff to triage by hand.
Best-in-class for Storybook component testing, with TurboSnap and a tight Storybook publish flow.
Tests the real app at real URLs, not just isolated stories — and classifies regression vs. intentional so review queues stay short. Storybook capture is supported too.
A free, self-hosted classic — simple, scriptable, no vendor at all.
Same self-hosted freedom, but with zero-config route discovery, anti-flake rendering, AI analysis, and active maintenance (BackstopJS has been quiet for years).
Modern, developer-friendly OSS-leaning tools with good CI ergonomics and Playwright trace support.
The only one with AI change classification and verified fixes — and a flat, screenshot-count-independent price with full self-hosting.
Frontguard reads your app by URL — no test files to port, no proprietary snapshot format. Point it at your dev server and you have baselines in one run. Migration guides walk through the rest.
Install the CLI and run your first AI-explained visual check in two minutes.