Status
Our own desks and every catalogued server, from the scheduled probe. A probe connects, lists tools, and runs one real call whose answer we check, because reachable and working are not the same thing.
Last run 2026-09-09 12:00 UTC.
Our desks
| Desk | Status | Protocol | Latency | Detail |
|---|---|---|---|---|
catalogue |
up | — | — | ok |
egf |
up | 2025-06-18 | 432 ms | ok |
egf-desk |
up | — | — | ok |
router |
up | — | — | ok |
sos |
up | — | — | ok |
utility |
up | — | — | ok |
Catalogued servers
| Server | Status | Score | Tools | Golden call | Detail |
|---|
How the score is read
| 0 | Unreachable, or refused the connection. |
| 1 | Connected, but the tool list failed or was empty. |
| 2 | Tool list fine, the golden call errored. |
| 3 | The call returned, but the answer was not what it should be. |
| 4 | Correct, but slow or with undocumented tools. |
| 5 | Correct, prompt, documented. |
One failed check is usually the network, so status stays up. Two consecutive is
degraded. Four, about a day, is down and drops out of catalogue search.
unknown means partner-gated with no public endpoint: we cannot check it, which is a
different claim from saying it is broken.