Open source and self-hosting
WebGuard’s code is public. You can read what the software does and run it on your own infrastructure.
- WebGuard Core (management UI, REST API, status pages, notifications) is released under the MIT license.
- The scanner instance that runs checks is a separate project, also MIT licensed.
- A Docker-based installation is available for self-hosting. You then decide where data is stored and which locations run checks.
Where checks run
The hosted version runs checks from monitoring locations that WebGuard lists publicly.
- One location is active today: de-1.
- The location list, with outbound IP addresses for allow-lists, comes straight from the WebGuard API and is shown on this site.
- Scanner instances authenticate to Core with their own key, and their metrics never include monitoring targets or credentials.
Privacy and tracking
Neither this website nor the WebGuard app uses analytics or marketing tracking.
- This website loads no third-party scripts, fonts, or images. A Content Security Policy limits scripts and connections to its own domain.
- It sets one cookie for your language and, only when you reveal the contact details in the imprint, a security cookie that expires after ten minutes. Your theme choice stays in your browser’s local storage.
- The WebGuard app sets only the cookies needed for sign-in, sessions, and CSRF protection. There is no cookie banner because there is nothing to consent to.
- Audit logs record user-controlled changes. Passwords, tokens, webhook URLs, HTTP bodies, and sensitive headers are redacted.
Authentication and access
- API keys are created with explicit abilities, such as “Read analytics” or “Write server health”. A key can only do what it was granted and can be revoked at any time.
- The MCP server requires a key with the “Read analytics” ability and returns only monitors your account may see. Its tools do not change monitors; only diagnose_incident with recompute stores a recalculated root-cause analysis.
- Teams have two roles, admin and member. Members get read-only access to shared team monitors; admins manage members and ownership.
- Sign-in, verification links, subscriptions, and API access are rate limited. API responses report the current limits in X-RateLimit headers.
Data retention
Scheduled jobs in WebGuard Core clean up data on a fixed schedule.
- Raw check results older than one week are moved weekly into an archive that keeps status, response time, and status code. Daily summaries are built every night and recalculate the last few days, so a missed run heals itself.
- Read notifications are deleted after one month, notifications of demo accounts after one week. Both jobs run daily.
- Audit log entries are cleaned up daily and deleted after 30 days.
- Deleted monitors are purged every month, together with their results, archive, SSL data, incidents, and notifications.
Transparency during incidents
- Public status pages show availability, uptime, incidents, and planned maintenance. Verified subscribers are notified by email.
- For every incident, WebGuard keeps the evidence behind the decision: the threshold and the fresh, stale, or missing result from each location.
- Automatic status updates stay neutral. Internal notes and diagnostics never appear on public pages.
- Root-cause analysis is rule-based and uses no external AI service. It shows the data behind every probable cause it lists. Configuration changes appear as field names only, never values, targets, or credentials.
How changes are verified
Every pull request runs automated checks in GitHub Actions. The details below come from the repositories’ workflows and documentation.
- WebGuard Core: Rector and Pint formatting checks, the full Pest test suite, and an OpenAPI contract check.
- Static analysis: PHPStan (Larastan) at level 8 over application code, configuration, and routes. Older findings are recorded in a baseline file; new and changed code may not add baseline entries or ignore comments.
- Test coverage: a coverage workflow requires at least 85% total coverage and 80% coverage of changed lines, and also runs every night.
- 15 browser tests (sign-in, onboarding, monitoring management, notification settings, responsive layout, and theme) run in a dedicated CI job in Chromium.
- A smoke test starts the container topology and opens a public status page in a real browser.
- Scanner instance: formatting check, go vet, Go tests including the race detector, govulncheck for known dependency vulnerabilities, and a Docker build on every change.
- This website: type checks, automated tests. Dependabot keeps dependencies current.
Report a vulnerability
If you find a security issue, please report it privately using the contact details in the imprint rather than opening a public issue. Describe what is affected and how to reproduce the behavior.
Go to the imprint