SaaS Launch Monitor

Most outages are found by a customer first. A page starts serving an error with a 200 status, a redirect breaks after a CDN change, a certificate stops renewing, or a health endpoint says the database is down while the homepage still loads.

SaaS Launch Monitor checks what "working" means for your product, from a small JSON file. Each check can test the status code (an expected 401 or 301 can pass), redirect chains including http to https, text that must or must not appear, fields in a JSON health response, response time, and the TLS certificate of every https host. Failures that look temporary are retried with backoff. Tokens come from environment variables and are redacted from every output.

In Claude Code, /launch-setup reads your homepage links, robots.txt and sitemap and proposes a config, then asks before saving. /launch-check runs the checks and explains the top issue in plain words. /launch-report turns the history file into uptime %, p95 latency and an incident list. /launch-incident drafts a customer status update and an internal incident note.

Inside: the plugin, a skill folder that also works on its own in OpenClaw, a Node.js checker with no npm packages, config and incident references, starter configs, an offline demo, and test records.

Limits: it runs when you or a scheduler call it. It is not a hosted service and sends no alerts itself. It checks raw HTML, so text added by browser JavaScript is not seen. It runs from one location.