Every release,
in the open.

Server load alerts are smarter about when to fire. A common false-positive pattern (high load average with low CPU usage, typical inside Linux containers and I/O-heavy workloads) is now correctly recognised as non-actionable and suppressed. Critical-tier load spikes still fire regardless of CPU so real incidents are not missed. The notification channel hint and load average tooltip on the server configuration dialog were also clarified.

• FixServer load alerts no longer fire for the common pattern of high load average with low CPU usage (typical of I/O-bound or containerized workloads). Critical-tier load spikes still fire as before.
• ImprovedClearer tooltip text on the server alert configuration dialog explaining when load alerts fire and where notifications get routed.

Internal-only fix for a regression in our continuous integration test suite. No customer-facing change.

• FixInternal build reliability fix.

When you accept an invitation to an existing organization, you now land directly on the dashboard. Previously, invited users were sent through the first-monitor setup wizard, even though the inviting organization already had monitors, channels, and a plan configured.

• FixInvited teammates joining an existing organization now skip the first-monitor setup wizard and land directly on the dashboard.

When a teammate accepted an invitation by signing up with Google or GitHub, the app was previously creating an extra empty workspace for them in addition to joining the inviting organization. They now land directly in the inviting organization with the correct role, no extra workspace.

• FixInvited teammates signing up with Google or GitHub no longer end up with an extra empty workspace in their account switcher.

Internal-only: we can now see why a signup or social login attempt did or did not complete, instead of only seeing the successes. No customer-facing change.

• ImprovedInternal telemetry improvements for the signup and social-login flow.

Public pages were loading two session-replay vendors. We removed Microsoft Clarity and kept the higher-quality one. Side effect: a known JavaScript error class disappears, public pages load slightly faster, and the cookie consent surface is simpler.

• ImprovedPublic pages now load one fewer third-party script for slightly faster first paint.
• FixResolved a recurring JavaScript error class that was firing on the homepage, sign-in, and status-page tools.
• ImprovedUpdated Privacy and Cookies pages to reflect the simplified vendor list.

Internal error monitoring no longer logs or forwards outgoing webhook URLs when paging engineering. No customer-facing change.

• SecurityEnhanced security hardening.

Internal log-level tuning so the engineering channel only pages on platform issues, not on customer-side connection failures that are already handled by the existing alert pipeline. No change to customer-facing alerts or check behavior.

• ImprovedInternal alerting noise reduction (no change to customer alerts).

When you click a notification from one organization while currently working in another, the dashboard now switches you to the correct organization automatically and lands you directly on the monitor, server, or incident the notification was about. No more 404, switch organization manually, then re-navigate.

• ImprovedNotification clicks across organizations now auto-switch you to the right organization and land you on the right page.

The AI-agent markdown rendering now leads with our actual differentiators (smart-alert clustering, Cloudflare integration, 15-minute visual regression, privacy-first status pages) instead of a generic feature list. Also fixed a delivery quirk where the homepage markdown was returning HTML when the prerender cache won the race against content negotiation.

• ImprovedAI-agent markdown rendering of public pages now actively pitches what makes us different rather than listing features.
• FixMarkdown rendering now reliably returns markdown for prerendered routes including the homepage.

Hotfix to the AI-agent markdown rendering shipped in the previous release: the homepage path now correctly returns markdown when requested with Accept: text/markdown.

• FixMarkdown rendering for AI agents now works on the homepage too (was returning HTML in 2.40.16).

Public marketing pages now also serve a clean markdown rendering when an AI agent or LLM tool requests it via the standard Accept: text/markdown header. Browsers continue to get the regular HTML view. Helps AI search engines and assistants ingest and cite Visual Sentinel content with higher fidelity.

• ImprovedPublic pages now respond with a clean markdown version when AI agents request it via Accept: text/markdown.

Fixed an intermittent rendering glitch on the homepage that some visitors were seeing on first load. Locked the live status pages against in-browser auto-translation that was causing the same glitch. Added our LinkedIn, Facebook, and Instagram presence to the footer and to our brand metadata so AI search engines can find them. Internal telemetry improvements for our signup funnel.

• FixResolved an intermittent rendering glitch on the homepage and live status pages.
• ImprovedAdded LinkedIn, Facebook, and Instagram links in the footer; expanded brand metadata for AI search visibility.
• ImprovedImproved internal funnel telemetry to better understand signup flow on different browsers.

The 29-day SSL expiry tier was retired across the dashboard because it fired during the typical certificate auto-renewal window (most customers got paged about a renewal that completed uneventfully). Existing monitors using the 29-day tier are migrated to the 20-day tier instead, which fires after auto-renewal would have completed and a real problem is more likely.

• ImprovedSSL expiry alerts no longer fire 29 days out by default. Replaced with a 20-day tier that fires after the typical auto-renewal window.
• ImprovedExisting monitors set to the 29-day tier are migrated to 20 days. All other custom tiers preserved.

Visual Monitoring now defaults to one capture per hour for new monitors, with a 15-minute minimum if you want it more frequent. Same alert quality, less noise, and lower load on the sites being monitored. Internal alerting noise from already-handled customer-side timeouts has also been reduced (your DOWN alerts are unchanged).

• ImprovedVisual Monitoring captures now default to once per hour for new monitors. Existing monitors keep their current cadence.
• ImprovedVisual Monitoring minimum is now 15 minutes (raised from 5).
• ImprovedInternal alert pipeline now distinguishes customer-side request timeouts from check infrastructure issues.

Tightened the check infrastructure trust model so a brief network blip on a check node no longer wakes you with a "site down" alert when your site is actually fine.

• FixShort-lived check infrastructure blips no longer trigger false-positive DOWN alerts.

Bulk actions on the alert settings page and the monitor list (linking monitors to a server) now apply to every selected monitor instead of silently failing on accounts with many monitors. The action cap now matches what the page can display.

• FixBulk alert and bulk server-linking actions now apply to every selected monitor.

Filtered out noise from a third-party library that cannot generate IDs on certain privacy-hardened browsers. No customer-facing behavior change.

• FixInternal error tracker filters a noisy false-positive class.

Sign-in via Google or GitHub now recovers automatically from a slow upstream call and gives up quickly with a clear retry path instead of leaving you waiting.

• ImprovedGoogle and GitHub sign-in is more resilient to slow third-party calls.

When you invite a teammate and they sign up directly (rather than clicking the invite link), they now land straight inside your organization with their assigned role. No detour, no orphan personal workspace, no extra step.

• ImprovedInvited teammates who sign up at the login page are added to your organization automatically.