How it works

Probes. Every 5 to 30 minutes, a Cloudflare Worker sends a tiny request ("Reply with the word ok") to each AI API with our own API key. We record the HTTP status, the error type and the latency.

User reports. Visitors click a button when a service does not work for them. We store the service, the kind of problem, the country and a hash of the IP address with a salt that changes every day. We never store the IP address itself. One report per service per 10 minutes counts.

Official status. We read the vendor status page every 10 minutes.

Verdict. "Problems" means the probe error rate is 5 times the 7-day normal, or user reports are far above normal, or the vendor reports degraded service. "Major outage" means a probe error rate of 30% or more, a very large report spike, or a vendor major outage.

Data. Raw data is deleted after 30 days. Hourly totals stay. JSON: /api/status.json. MCP server: https://aidown.dev/mcp.