Measured
every 60 seconds.
A synthetic probe checks every hosted team server every 60 seconds. This page renders those raw aggregates: the same numbers that gate our own pricing decisions. If a day was red, you will see it here.
The last 30 days.
A day is green when probe success holds ≥ 99.5% and every server-reported signal stayed inside its objective. Missing signals never count as green; absence of evidence is shown as exactly that.
Probe success
A synthetic probe hits every hosted team server every 60 seconds from outside.
p95 latency
Server-reported rolling p95 across all /v1 requests.
Availability
Server-reported rolling availability over the last 30 days.
Index freshness
Worst lag between a pushed change and its visibility in hosted search.
GA gate: 30 consecutive green days
We hold ourselves to the same bar we publish: the hosted-index price move (roadmap #374) only ships after 30/30 green days on this page. Verdict right now: —
Where these numbers come from.
- A probe in the control plane calls
GET /v1/metricson every hosted team server every 60 seconds: outside-in, through the same edge your requests take. - Every result is stored raw (success, latency, server-reported p95 / availability / index lag). Nothing is filtered before aggregation.
- This page fetches the anonymous fleet aggregate (raw JSON, 5-minute cache). No account data, no server URLs, no cherry-picking.
- The objectives and the green-day rule are versioned in the open repository (team-slos.toml): the same definitions the engine enforces locally.
Local lean-ctx never depends on any of this: the engine, cache, ledger and memory are local files on your machine. This page covers only the optional hosted team plane.
Want this rigor for your own agents?
The same SLO engine ships in the open core: define objectives in team-slos.toml and lean-ctx enforces them locally, no cloud required.