System status.
The figures below are real and revalidate hourly. Authoritative uptime is monitored externally on an independent status page that stays up when this site does not; the endpoints are public, so you can check liveness yourself rather than take a green light on faith. Failure behavior per surface is documented at /reliability.
- Registry snapshotsynced several times a day from the official MCP registry
- just now
- Servers tracked
- 25,774
- Verdicts on filecoverage expands adversarial-first; see /methodology
- 24,105
/api/registry-countlive server count (JSON)/api/v1/trust/tool/<server_id>/<tool_name>verdict API · free tier, no key/api/v1/screenlive screener (POST)
The hosting platform disabled all deployments after the account crossed a compute usage cap; every page and API on this domain returned HTTP 402 until the plan was upgraded. Detected by monitoring in 6 minutes; resolved in 3h40m. No data was lost, no verdict changed, and locally installed gates kept enforcing their pins throughout - the failure boundary held as documented at /reliability. Follow-ups shipped: an independent status page, account usage alerting, and a scoped write credential for the verdict pipeline.
Failures stated plainly are part of the same honesty contract as the verdicts.
More numbers: how many MCP servers there are · trust model