Every release,
in the open.

Tightened the rules that decide when a slow-response alert fires, so a brief blip during a check no longer trips the alert. Your configured slow-response threshold is now respected exactly as set, and slow checks are counted per region instead of pooled across regions, which removes a noise pattern where a single transient hiccup could fire alerts for several monitors at once.

• FixSlow-response alerts now require the full number of consecutive slow checks you set on the monitor before firing, instead of half that count.
• FixSlow-check streaks are now counted per check region, so a brief transient slowness across multiple regions no longer counts as multiple slow checks against the same monitor.
• ImprovedRecovery alerts that paired with these false slow-response incidents will also stop, giving you a quieter alert inbox.

Added a new side-by-side comparison page under /compare. Also published the public OpenAPI specification on GitHub so developers and AI agents can integrate with the Visual Sentinel API directly from a canonical source.

• NewNew comparison page added under /compare.
• NewPublic OpenAPI specification published at github.com/VisualSentinel/openapi with curl, JavaScript, and Python quick-start examples.

New blog posts and changelog entries now notify search engines the moment they go live, so they show up in search results in hours instead of days. Previously, a freshly published post had to wait for the next natural crawl cycle.

• ImprovedNew blog posts are pushed to indexing endpoints automatically on publish.
• ImprovedTidier brand metadata across the site.

The Add Monitors and Add Servers dialogs on a status page now have a search box and a Select All checkbox. Search by name or URL, then click Select All to pick the entire matching set in one go. The previous one-click-per-row workflow still works for picking individual entries.

• NewAdd Monitors / Add Servers dialog now has a search box and a Select All checkbox so you can bulk-link many entries to a status page in one click.
• ImprovedSelect All operates on the currently filtered set, so typing a search term then clicking Select All picks just the matching entries.

If you manage many monitors, the Monitors page used to slow down once the list grew large because every card was rendered up front. The list and grid views now render only what is on screen, so scrolling, searching, and filtering stay snappy regardless of how many monitors you have. Smaller accounts continue to render exactly as before.

• ImprovedMonitors page list and grid views now render only the cards visible on screen, keeping scrolling and filtering smooth on large accounts.

Bespoke Enterprise customers were being capped at the standard Agency monitor limit even when their account carried a higher per-customer ceiling. The plan-limits resolver now correctly honors per-customer overrides for Enterprise accounts.

• FixEnterprise customers with custom monitor / team / status-page / data-retention ceilings on their account now have those limits enforced correctly at gate-check time, instead of falling back to standard Agency limits.

Status pages now support two access controls. Add a password to protect any status page, set an IP allowlist to restrict access to specific IP addresses or ranges, or combine both. The IP allowlist requires a verified custom domain. Existing public status pages render exactly as before unless you turn on a control in Settings.

• NewPassword-protect any status page with a single password that visitors enter to view it.
• NewIP allowlist for status pages on a verified custom domain. Add IPv4 or IPv6 addresses, or CIDR ranges. Visitors outside the allowlist see an access-denied message.
• NewBulk-import IPs by pasting a comma, semicolon, space, or newline separated list, or by uploading a CSV / TXT file (up to 200 KB). Lines starting with # are comments.
• NewAdd IPs one at a time with inline validation, plus a removable list view of every entry.
• ImprovedNew Privacy & Access section in Status Page Settings ties everything together. Both controls can be combined.

Several feature gates were checking your subscription tier without taking your active trial into account, so trial customers were silently denied features they were entitled to during the trial. The DNS history snapshots, keyword content monitoring, status-page theme and SEO controls, and the AI usage allowance on the billing page all now honor the trial state correctly. Once your trial ends, normal plan limits apply as before.

• FixDNS history snapshots are now captured on schedule for trial customers, so the DNS tab populates without needing to open each monitor manually.
• FixKeyword and content-change monitoring now works during your trial without being silently downgraded to the Free-plan feature set.
• FixStatus-page theme, custom CSS, custom HTML, SEO meta fields, and OG image are editable during the trial window for plans that include those tiers at conversion.
• FixAI usage page on billing now shows the correct token allowance during your trial instead of the Free-plan zero allowance.
• ImprovedInternal audit added a recurring guard so any future feature gate that reads your plan also reads your trial state.

When a monitor wasn't actively checking during a report period (because it was disabled, paused, or newly created with no checks yet), reports used to show '100% uptime' on the headline stats. They now correctly show 'N/A' and the AI summary explains that monitoring was inactive instead of leading with a misleading score. Reports for active monitors are unchanged.

• FixReports for monitors that were inactive during the report window now display N/A for uptime, response time, and success rate, instead of a fabricated 100%.
• ImprovedAI report summaries no longer claim "achieved 100% uptime" for inactive monitors and now flag the monitoring gap as a configuration issue.
• ImprovedOrganization-wide weekly digest averages now exclude inactive monitors instead of letting their fabricated scores skew the rollup.

Tightened internal error reporting so common bot-probe traffic against the platform's homepage no longer surfaces as on-call alerts. No customer-facing change.

• ImprovedInternal alerting tuned to suppress noise from bot probes without affecting genuine error visibility.

If you started your trial with a card on file, the billing page now correctly shows your billing-start date and tells you can cancel anytime before then, instead of the generic 'subscribe below to keep your features' message. Customers without a card on file continue to see the 'subscribe to keep features' message as before.

• FixCard-on-file trials now show the correct billing-start date and cancellation language on the billing page.

Recovery (back online) alerts now fire when an incident actually closes, rather than relying on a fixed time window. This guarantees recovery alerts for genuine multi-day outages while still suppressing alerts from background state cleanup or brief check-time flaps that never crossed the alerting threshold.

• FixRecovery alerts now reliably fire for genuine multi-day outages, including those at the maximum repeat-alert cadence.
• ImprovedRecovery alerts are now keyed on the underlying incident lifecycle, so brief check-time flaps that did not generate a down alert no longer produce a recovery alert either.

Monitor recovery alerts now require a corresponding down alert within the last 24 hours. This prevents 'recovered' notifications from firing when no outage was ever surfaced to you, which could otherwise happen during background state cleanup after platform upgrades.

• FixRecovery alerts will only fire when there was a corresponding down alert in the last 24 hours, eliminating noise from background state-machine cleanup.

Down detection for monitors that check from multiple regions has been hardened. A subtle interaction between the database server's clock-zone setting and certain time-window queries could prevent the platform from recognising that recent checks had failed across regions. The fix is layered across the application code, the database connection, and the runtime environment so the issue cannot recur as we add more check regions and nodes.

• FixDown detection for multi-region monitors is now reliable regardless of the database server's clock-zone configuration.
• ImprovedHardened the platform end-to-end against clock-zone drift as we expand to new monitoring regions, so adding new check nodes cannot reintroduce this class of issue.

If your monitor went down and you got a 'monitor is down' alert, you should always get the matching 'monitor recovered' alert when it comes back. A subtle gap in the previous logic could occasionally drop the recovery alert when an outage resolved very quickly or affected only one region. Recovery alerts are now tied directly to the incident lifecycle, so every closed incident that originally alerted you also tells you when it's resolved.

• FixRecovery (back online) alerts now reliably fire whenever a previously-alerted incident resolves, including fast outages and single-region issues that did not always trigger a recovery alert before.
• ImprovedRecovery alerts are now strictly one-per-incident, so duplicate recovery messages are not possible even if multiple internal paths agree the monitor is healthy at the same time.

When a monitor had no explicit visual check cadence configured, the settings UI was showing 'Every 24 hours' regardless of the actual cadence. The display now reflects the cadence the system is really using for that monitor, taking your plan into account. Saving the form without changing the dropdown no longer writes a value, so your intent to 'use the plan default' is preserved across plan changes. No customer action required.

• FixMonitor settings now show the actual visual check cadence the system uses, instead of a generic 24-hour placeholder when no explicit value was set.
• ImprovedSaving visual settings without changing the cadence dropdown no longer locks in a value, so the cadence continues to follow your plan defaults if you upgrade or downgrade later.

We rewrote earlier changelog entries to be cleaner and more user-focused. Same shipped behavior, tidier copy.

• ImprovedRefreshed the wording on a number of historical changelog entries.

Background cleanup of stale visual-monitoring records now runs reliably. A pre-existing bug was causing the cleanup pass to abort early on high-traffic accounts, which over time could allow stale records to accumulate. No customer action required.

• FixBackground cleanup of stale visual-monitoring records now runs reliably on every restart. Prevents accumulation on high-traffic accounts.

Visual change detection now requires the same content drift to repeat across multiple checks before triggering a screenshot capture. Previously, any mismatch between checks counted toward the threshold, which could falsely trigger captures on pages with dynamic noise (timestamps, ad slots, dynamic markup) even when no real visual change had occurred. With this release, random per-request noise resets the counter; only stable, repeated changes lead to a screenshot. Once a screenshot is taken for a particular changed state, that exact state is suppressed from re-capture until either the page recovers to baseline or the periodic safety window elapses (default 12 hours, configurable). The result is dramatically fewer wasted captures on dynamic pages and a much higher signal ratio in alerts. No customer action required.

• ImprovedVisual change detection now requires the same content drift to repeat before triggering a screenshot. Random per-request page noise no longer accumulates toward false-positive captures.
• ImprovedOnce a screenshot is captured for a particular changed state, that exact state is suppressed from re-capture until the page recovers or the periodic safety window elapses (default 12 hours).

Foundation work for running visual monitoring at much higher monitor counts. Adds centralized rate limiting (per-minute global and per-host), smart prioritization (manual checks beat scheduled ones), backpressure-aware delay when the queue is hot, and deduplication so the same capture can't queue twice. Also adds an opt-in scanning mode that replaces per-monitor scheduled jobs with a single periodic scan, drastically lowering background overhead. The new mode is opt-in; existing accounts continue to use the previous behavior. No customer action required.

• ImprovedVisual capture pipeline now enforces global and per-host rate limits with smart prioritization, deduplication, and queue-aware backpressure. Reduces wasted captures and improves responsiveness under load.
• ImprovedNew opt-in eligibility scanner replaces per-monitor scheduled jobs with a single periodic scan, foundational for running visual monitoring at much higher account sizes. Existing behavior is unchanged unless explicitly switched on.
• ImprovedVisual change detection threshold can now be configured per monitor (including zero for instant-trigger), and respects the configuration on slow monitors that previously overrode it.