Measured uptime
Every row below is counted from our own checks — one per route per hour, stored and never edited. The weakest route is first, because a status page that only shows green is an advertisement.
Window: 15 Sept 2026 to 22 Sept 2026 (7 days).
| Route | Answered | Failures | Median | p95 |
|---|---|---|---|---|
medium/profile | 75.9% of 166 asks | 40 | 4,502 ms | 11,822 ms |
instagram/profile | 95.6% of 159 asks | 7 | 6,248 ms | 11,798 ms |
tiktok/profile | 98.7% of 159 asks | 2 | 1,826 ms | 6,342 ms |
bluesky/posts | 99.4% of 166 asks | 1 | 959 ms | 3,059 ms |
bluesky/profile | 99.4% of 158 asks | 1 | 1,000 ms | 2,802 ms |
linktree/profile | 99.4% of 166 asks | 1 | 934 ms | 3,060 ms |
github/profile | 100% of 158 asks | 0 | 2,054 ms | 3,650 ms |
mastodon/posts | 100% of 166 asks | 0 | 1,672 ms | 5,373 ms |
mastodon/profile | 100% of 166 asks | 0 | 1,031 ms | 3,827 ms |
pinterest/profile | 100% of 157 asks | 0 | 1,643 ms | 4,755 ms |
soundcloud/profile | 100% of 158 asks | 0 | 1,165 ms | 3,486 ms |
threads/profile | 100% of 166 asks | 0 | 1,852 ms | 6,159 ms |
How this is counted
A check counts as an ask only when it actually reached the platform. Cache hits are left out of the count entirely — they are neither a success nor a failure, because no question was asked. Counting them as successes would inflate every row; counting them as failures would understate them.
Median and p95 describe successful responses only. A refusal is not a response, and letting refusals into the latency figure would answer a different question than “how long does it take to answer”. The percentiles are observed values, not interpolations: every number on this page is a measurement that happened.
About the weakest row
Medium refuses our addresses in windows that last one to four hours and clear on their own. Two things happen when it does: the monitor retries on a different handle in the same round, and — if you ask for it with Cache-Control: stale-if-error — the API serves the last stored reading with an explicit as_of instead of an error. Both are described in the documentation.
What this page does not claim
This table covers 12 of the 19 routes we sell. The other seven received no checks at all in the same seven days: account, archive/coverage, archive/movers, archive/trends, github/repos, profile/history and threads/posts. Counted on 22 September 2026. Every route in the table above costs 1 credit; the seven missing ones cost up to 10.
Missing from this page does not mean unreliable. It means unmeasured: we are not saying those routes failed, and we are not saying they answered. How the checked set came to be the set it is — and the one query that finds the same hole in any API — is in 284 green checks, and the expensive half was never checked.
Seven days is not a quarter, and roughly 150 rounds per route is one week of sample. These numbers say what happened in that window — not what will happen next month. They are also our measurements of our routes, taken from one network; a check from somewhere else could see something different.
Looking for the trend archive instead? Its health is at /status.