Uptime monitoring for the AI era
Uptime, synthetic, and AI-agent monitoring in one tool: HTTP/TCP/cron checks, LLM-judged agent checks, run observability, instant alerts, and status pages. Set up in 60 seconds.
Features
Monitor web endpoints, raw ports, host reachability, and background jobs with one tool. Keyword matching and SSL-expiry checks included.
Prompt your agent and let Claude judge whether the reply is actually correct — catching 'up but giving garbage,' not just HTTP 200. Test a check live before you save it.
Drop in our SDK and every agent run reports its latency, token usage, cost, and tool steps. Per-agent dashboards with error-rate and cost trends.
Set thresholds on error rate or cost and get paged the moment an agent degrades or starts burning money — routed to the channels you choose.
Email, Slack, and webhooks fire only on confirmed up/down transitions — not on every flaky probe. Configurable failure thresholds.
When something breaks, Claude reads the response and error data and tells you the likely cause and fix — no log spelunking.
Share a branded status page with your users in one click, complete with 90-day uptime history.
Paste your app's URL and we crawl it to propose monitors, then re-scan daily to flag new endpoints as you ship.
Check from multiple regions so you catch geography-specific outages and avoid false alarms.
FAQ
AI-Monitring checks HTTP and HTTPS endpoints (with status-code, keyword and SSL-expiry validation), raw TCP ports, host reachability via ICMP ping, background/cron jobs via heartbeat monitoring, and AI agents — both with synthetic LLM-judged checks and with run telemetry reported from your code.
Two ways. Synthetic agent checks send a prompt to your agent endpoint on a schedule and use Claude as a judge to decide whether the response is correct — catching 'up but giving wrong answers,' which a normal HTTP check misses. Agent observability is the inside-out view: install the @ai-monitring/sdk package and each agent run reports its latency, token usage, cost, tool steps, and success/failure to per-agent dashboards.
Yes. Set threshold rules on error rate or cost — per agent or across all agents — over a time window. When a threshold is breached you're alerted through your chosen email, Slack, or webhook channels, with a recovery notification when it clears.
Check intervals range from every 30 seconds on the Business plan to every 5 minutes on the Free plan. You choose the interval per monitor.
Via email, Slack, and generic webhooks. Alerts fire only on confirmed up/down transitions using a configurable failure threshold, so you don't get paged for a single flaky request.
AI-Monitring needs a Slack Incoming Webhook URL (it looks like https://hooks.slack.com/services/...), not the channel's app.slack.com browser link. Create one at https://api.slack.com/apps → Create New App → From scratch, pick your workspace, open Incoming Webhooks and turn it on, then Add New Webhook to Workspace and choose the channel you want alerts in. Copy the generated hooks.slack.com URL, then in AI-Monitring go to Dashboard → Alert channels, add a Slack channel and paste it. Click Test to send a sample alert and confirm it arrives.
When a monitor goes down, Claude reads the HTTP response, status code, and error data and produces a plain-language summary of the likely cause and a suggested fix, so you spend less time digging through logs.
Yes. Every plan includes hosted public status pages with a shareable URL and 90-day uptime history. You select which monitors appear and customise the title and branding.
Yes. The Free plan includes 5 monitors, 5-minute intervals, one status page, and email alerts — no credit card required.
AI-Monitring runs on NEXUS AI with a managed PostgreSQL database, deployed via the NEXUS AI CLI.