Live probes
MCP Network Health· skil.ski
Automated hourly probes across every federated MCP and SaaS skill in the catalog. Degraded and down skills are route-arounded in default search results.
Last updated: May 4, 08:17 PM UTC
Healthy
0
Degraded
0
Down
0
P50 Latency
—
Total Probed
0
Unknown
0
How we monitor
Every hour, an automated job probes up to 50 federated MCP and SaaS skills from the catalog using a HEAD (falling back to GET) request with a 5-second timeout.
- Healthy — HTTP 2xx response in under 2 000 ms.
- Degraded— HTTP 2xx but slow (≥2 000 ms), or 4xx non-auth response.
- Down — HTTP 5xx, connection refused, or probe timeout.
- Unknown — Not yet probed.
Degraded and down skills are penalized in the default search sort so agents are routed to healthy alternatives first. This drives Search-to-Fill Rate toward 99%.
API access
curl skill.ski/api/registry/health | jq
CORS-open · JSON · max-age=300
Probe results reflect external HTTP reachability only, not internal skill execution quality. A healthy probe confirms the endpoint is reachable; it does not certify tool output correctness. For execution quality data, see Search-to-Fill Rate.