- 99.95%Uptime · 30 days
- 483 msp50 response time
- 1.74 sp95 response time
Right now
All systems operational
Trailing windows
Two windows. Both dated.
The 30-day window is what this page reports in detail. The 90-day window is the one the figures elsewhere on this site quote, kept here so the two can be compared against the same method.
| Figure | 30 daysJun 29 – Jul 28, 2026 | 90 daysApr 30 – Jul 28, 2026 |
|---|---|---|
| UptimeMean of every surface | 99.95% | 99.95% |
| Response time, p50The median call | 483 ms | 496 ms |
| Response time, p9595 of 100 calls come in under this | 1.74 s | 1.79 s |
| Response time, p99The tail | 2.90 s | 3.01 s |
| Mean response timeArithmetic mean of every check | 617 ms | 641 ms |
| Failed minutesSummed across surfaces | 164 | 520 |
Uptime by surface
8 surfaces, checked separately.
Each surface is probed on its own, so an incident on one is never averaged away by the others. A day is marked down only when checks actually failed.
How to read a day
- Every check passed
- Slow, still answering — not counted as downtime
- Checks failed — minutes counted against uptime
| Surface | Uptime | 30 days | Status |
|---|---|---|---|
| Profile APIProfile enrichment, experience, education, and activity. | 100.00%no failed minutes | 29 operational days, 1 degraded, 0 with an outage. | Operational |
| Company APICompany records, headcount, and job counts. | 100.00%no failed minutes | 30 operational days, 0 degraded, 0 with an outage. | Operational |
| Search APIPeople, company, and location search with filter discovery. | 100.00%no failed minutes | 30 operational days, 0 degraded, 0 with an outage. | Operational |
| Jobs APIJob search, job detail, and hiring signals. | 99.62%164 failed minutes | 29 operational days, 0 degraded, 1 with an outage. | Operational |
| Posts APIPosts, comments, and reaction data for public activity. | 100.00%no failed minutes | 30 operational days, 0 degraded, 0 with an outage. | Operational |
| AuthenticationAPI key validation, quota checks, and rate limiting. | 100.00%no failed minutes | 30 operational days, 0 degraded, 0 with an outage. | Operational |
| MCP serverThe Model Context Protocol server over the same key. | 100.00%no failed minutes | 30 operational days, 0 degraded, 0 with an outage. | Operational |
| DashboardKey management, usage, logs, and billing. | 100.00%no failed minutes | 30 operational days, 0 degraded, 0 with an outage. | Operational |
| All surfaces · unweighted mean | 99.95%164 failed minutes | Jun 29 – Jul 28, 2026 · 99.95% over 90 days | |
Response time
Three percentiles, not one average.
End-to-end time measured at the probe, including TLS and transfer, over the last 30 days. A single average hides the tail that integrations actually feel, so the median, the p95, and the p99 are all published.
The published figure
1.80 s
Every bar below is drawn against this number: a full track is a p95 that has reached the response time quoted elsewhere on the site. Pooled across all API checks, the 30-day p95 is 1.74 s and the mean is 617 ms.
| Surface | p50 | p95 | p99 | p95 against 1.80 s |
|---|---|---|---|---|
| Profile API | 431 ms | 1.61 s | 2.48 s | p95 is 89% of the published 1.80 s figure. |
| Company API | 488 ms | 1.72 s | 2.74 s | p95 is 96% of the published 1.80 s figure. |
| Search API | 664 ms | 1.78 s | 3.12 s | p95 is 99% of the published 1.80 s figure. |
| Jobs API | 470 ms | 1.69 s | 2.61 s | p95 is 94% of the published 1.80 s figure. |
| Posts API | 502 ms | 1.75 s | 2.88 s | p95 is 97% of the published 1.80 s figure. |
| Authentication | 74 ms | 210 ms | 340 ms | p95 is 12% of the published 1.80 s figure. |
| MCP server | 551 ms | 1.76 s | 2.96 s | p95 is 98% of the published 1.80 s figure. |
| Dashboard | 214 ms | 640 ms | 980 ms | p95 is 36% of the published 1.80 s figure. |
| All API checks · pooled | 483 ms | 1.74 s | 2.90 s |
Methodology
How every figure here is produced.
A number without a method is a claim. This is the method, stated in enough detail that you could disagree with it.
- What is checked
- 8 surfaces, one synthetic check each, calling a documented endpoint with a fixed request and a stable identifier. The checks run against production, on the same paths and the same authentication that customer traffic uses.Browse the API reference
- Where from
- Three regions — us-east, eu-west, ap-southeast. A minute is recorded as failed only when the check fails from at least two of the three, so a single region's network trouble is not published as an EnvoAPI outage.
- How often
- Every 60 seconds per region: 43,200 checks per surface per region across a 30-day window.
- What counts as downtime
- A check that returns a 5xx, fails to connect, or exceeds the 30-second timeout. Client errors caused by the request itself — a 400, a 401 on a revoked key, a 404 for an identifier that does not exist — are not counted, because the API answered correctly.Read the error reference
- What counts as degraded
- Checks succeed, but the surface's p95 stays above twice its trailing baseline for 15 minutes or more. A degraded day is marked on the daily record above; those minutes are never subtracted from uptime, because no call failed.
- How uptime is computed
- Per surface: 1 − failed minutes ÷ minutes in the window. The site-wide figure is the unweighted mean of the 8 surfaces — not a traffic-weighted average, which would let a busy healthy surface mask an outage on a quieter one.
- What the percentiles mean
- End-to-end response time measured at the probe, including TLS and transfer. p50 is the median call; p95 is the value 95 of 100 calls come in under; p99 is the tail. Window percentiles are computed from the pooled sample of every check, never averaged from daily percentiles.
- Where these figures appear
- The uptime and response-time figures quoted on the homepage and the API pages are the 90-day numbers from this page: 99.95% uptime and a 1.79 s p95 across all API checks, measured Apr 30 – Jul 28, 2026.
- Maintenance
- Minutes inside an announced maintenance window are excluded from uptime only when the surface is actually taken offline; work that keeps the surface serving is not excluded. Unannounced work is never excluded.
- Corrections and retention
- Daily records are kept for 90 days. Figures are recomputed once per day at the collection time above; a figure corrected after publication is republished with a note rather than replaced silently.
- Not a service level
- These are historical measurements of what happened, not a commitment about what will. EnvoAPI does not offer uptime commitments, guaranteed response times, or service credits unless an Order Form says so.Support and service levels, Terms §13
- Data source
- Provisional. This page publishes the reporting format and the method above while the live probe feed is being connected; the figures shown are sample data and are not yet measurements of production traffic.