Pricing that respects
open source.
Run the local CLI for free without an account. Hosted plans are not generally available; the waitlist is for design-partner conversations, not purchase.
The full CLI. Everything you need to catch visual bugs in CI.
npm install @frontguard/cli- ✓Unlimited screenshots & routes
- ✓Multi-browser & multi-viewport
- ✓AI analysis (bring your own key)
- ✓Experimental fix generation + opt-in verification
- ✓Git-native baselines
- ✓Generated CI workflow + HTML artifact
- ✓All 5 built-in plugins
Pre-release hosted concepts for teams willing to help validate the workflow.
Join the waitlist- ✓Candidate hosted dashboard & report history
- ✓Candidate managed baseline storage
- ✓Candidate team notifications
- ✓Reference rendering under evaluation
- ✓Design-partner feedback channel
A roadmap discussion, not a generally available enterprise service or SLA.
Contact us- ✓Candidate team roles & invitations
- ✓Candidate baseline review workflows
- ✓Candidate activity and audit surfaces
- ✓No production billing flow today
- ✓No production SSO flow today
- ✓No support SLA today
Cloud source is included for evaluation, but the hosted and self-host onboarding paths are pre-release and do not carry a production support commitment.
Compare plans
| CAPABILITY | Open Source | Hosted (planned) | Team (planned) |
|---|---|---|---|
| CLI: render, diff, report | ✓ | ✓ | ✓ |
| AI analysis (BYOK) | ✓ | ✓ | ✓ |
| AI fix generation & verification | ✓ | ✓ | ✓ |
| Hosted dashboard & history | Local only | Planned | Planned |
| Managed baseline storage | Git | Planned | Planned |
| Production monitoring scheduler | CLI | — | Planned |
| Slack / PagerDuty alerts | Webhook | Planned | Planned |
| Teams, roles & approvals | Local only | Single-user | Planned |
| SSO & audit log | Local only | Single-user | Planned |
Questions
How do I install Frontguard?
Run npm install @frontguard/cli, initialize the config, start your app, and review it before running frontguard update-baselines --confirm-baselines. Push the frontguard-baselines branch before CI comparisons. The Playwright adapter installs with npm install -D @frontguard/cli @frontguard/playwright.
How does Frontguard handle cross-OS rendering differences?
Playwright rendering can vary by OS and hardware. The renderer is currently repository-source-only and requires the documented npm pack/tarball preparation before frontguard run --docker; cross-host byte equivalence has not yet been validated and no image is published.
Can I self-host the cloud?
The MIT-licensed cloud source is available for evaluation, but its Docker Compose path is not a verified clean-checkout deployment and no supported self-host quick start or SLA exists yet.
What environment variables does Frontguard read?
For AI: FRONTGUARD_OPENAI_KEY, FRONTGUARD_ANTHROPIC_KEY, or FRONTGUARD_GEMINI_KEY (plus FRONTGUARD_OPENAI_BASE_URL for custom OpenAI-compatible endpoints). The Playwright plugin also accepts unprefixed OPENAI_API_KEY / ANTHROPIC_API_KEY / GEMINI_API_KEY when present. MCP and cloud-development tools support local report mode via FRONTGUARD_LOCAL_REPORT or require an explicit FRONTGUARD_API_URL and FRONTGUARD_API_KEY; there is no live hosted default. frontguard doctor reads the same AI names as the runtime.
OpenAI or Anthropic — which should I use?
Choose the provider in frontguard.config.ts and, optionally, the model: OpenAI (default gpt-6-luna), Anthropic (default claude-sonnet-5-5), Google Gemini (default gemini-3.5-flash-lite), and custom OpenAI-compatible baseUrl endpoints (OpenRouter, Ollama, vLLM) are all supported. Run npm run eval:ai to compare a model on your machine before switching. When AI is enabled, Frontguard sends screenshot evidence, basic route metadata, and optional accessibility findings directly to that provider; it does not send a DOM snapshot or console log.
Does Frontguard work with Storybook?
Yes. frontguard init detects an existing Storybook (looks for .storybook/main.ts) and scaffolds a Storybook-aware config. The adapter walks the Storybook iframe, runs each story's play() function, and produces one screenshot per story by viewport.
Is there an MCP server for in-IDE agents?
@frontguard/mcp exposes 8 tools for coding agents: list_regressions, get_suggested_fix, recent_runs, accept_baseline, inspect_diff_cluster_dom, verify_css_patch, list_horizontal_overflow, and promote_baseline. It works locally against ./frontguard-report/report.json (or FRONTGUARD_LOCAL_REPORT) without a server, or against an explicit FRONTGUARD_API_URL deployment.
What's the data retention policy for screenshots?
The local CLI keeps baselines in git and reports on disk unless you configure AI or image upload. The pre-release cloud has no production retention commitment: general run images are not automatically pruned, team deletion attempts object cleanup on a best-effort basis, and operators must reconcile failures.
Start with the free local CLI.
Install the CLI without an account, start your app, explicitly accept reviewed baselines, then run the comparison again.