Is Jev Down? TypeSafe API Status and Outage History
Last checked · Independent guide, not affiliated with TypeSafe AI
To check whether Jev is down, open TypeSafe's official status page, status.typesafe.ai. On September 19, 2026 it showed the TypeSafe API as operational with 99.854% availability over the previous 90 days; the most recent downtime it recorded was 5 minutes on September 17. If the status page is green, the problem is most likely in your request or your channel.
Where to check
Section titled “Where to check”| Source | What it tells you |
|---|---|
| status.typesafe.ai | TypeSafe’s official status page: current state of api.typesafe.ai, 90 days of daily uptime, incidents and maintenance |
| A test call | Whether your key, network and request work right now (see below) |
| Your channel’s status | If you use Vercel, OpenRouter or Cloudflare, their own platform status matters too |
What the status page showed (checked September 19, 2026)
Section titled “What the status page showed (checked September 19, 2026)”- Current state: all services online;
api.typesafe.aioperational. - Availability over the past 90 days: 99.854%.
- History: the record goes back to June 21, 2026, three months before the public launch, which suggests the API was already serving private users.
Downtime recorded since the start of August:
| Date | Recorded downtime |
|---|---|
| Aug 3, 2026 | 3 minutes |
| Aug 4, 2026 | 59 minutes |
| Aug 6, 2026 | 2 minutes |
| Aug 22, 2026 | 8 minutes |
| Sep 3, 2026 | 3 minutes |
| Sep 8, 2026 | 3 minutes |
| Sep 10, 2026 | 6 minutes |
| Sep 17, 2026 | 5 minutes |
Launch day, September 15, shows no recorded downtime, although TechCrunch reported that demand briefly left TypeSafe unable to serve users from its API that week. Status pages measure their own checks; short capacity problems can show up as 529 errors for some users without appearing as downtime.
Is it Jev or is it you?
Section titled “Is it Jev or is it you?”Run the smallest request that exercises your key:
curl -s -w "\nHTTP %{http_code} in %{time_total}s\n" https://api.typesafe.ai/v1/models \ -H "Authorization: Bearer $TYPESAFE_API_KEY"Then read the result:
| You get | Most likely |
|---|---|
| HTTP 200 with a list of models | The API is up. Look at your request: see the errors index |
| 401 or 403 | Your key or header, not an outage. See 401 and 403 |
| 429 | You are over your rate limit. See 429 |
| 529 | TypeSafe is overloaded. Retry with backoff. See 529 |
| Timeouts or connection errors while the status page is green | Often your network or a proxy; try from another network |
| 5xx errors and a red status page | A real outage: wait, and fall back if you can |
In our tests from East Asia, healthy calls returned in 250 to 600 ms. Latency far above that on small requests is worth investigating even without errors.
If you call Jev through another platform
Section titled “If you call Jev through another platform”Vercel AI Gateway, OpenRouter and Cloudflare Workers AI forward requests to the same model, so a TypeSafe outage affects them too. But they can also fail on their own: account problems (a missing card on Vercel, no credits on OpenRouter), their own rate limits, or their own incidents. Check the platform’s status page as well as TypeSafe’s.
Build for the next outage
Section titled “Build for the next outage”Every hosted API has bad minutes. For anything user-facing:
- Set a timeout and a fallback path, such as a default route, a human queue or a rule.
- Let the SDK retry 429 and 5xx responses with backoff (both official SDKs do by default).
- Queue batch work so failed items are retried later instead of lost.
We re-check the status page when we update this page. The “Last checked” date at the top shows when that was.