Is the co.civai.nova/mcp-juicyway MCP server working?

Is the co.civai.nova/mcp-juicyway MCP server working?

No, not as the registry describes it. 22 declared endpoints failed the MCP initialize handshake on 2026-11-02 — the protocol's own way of asking a server to identify itself. No tool was ever called.

ERR-2026-8417, ERR-2026-8418, ERR-2026-8419, ERR-2026-8420, ERR-2026-8421, ERR-2026-8422, ERR-2026-8423, ERR-2026-8424, ERR-2026-8425, ERR-2026-8426, ERR-2026-8427, ERR-2026-8428

Why can't I reach the co.civai.nova/mcp-juicyway API?

Its declared address answered neither a request nor a TLS handshake on 2026-11-02, across 2 cases. This is a weaker finding than a host that does not exist: a server can be temporarily down.

ERR-2026-8434, ERR-2026-8438

Every record behind those answers

24 open records, observed 2026-09-10. Every claim below carries the command that disproves it. If one is wrong, it should be withdrawn — 1184 have been across the whole registry.

the endpoint rejects a mandatory part of the handshake (22)

The declared endpoint accepts initialize but rejects the mandatory notifications/initialized that follows it, with HTTP 400. A client following the specification breaks here.

Check it yourself:

curl -s -X POST https://nova.civai.co/mcp/agents/mcp-android_developer -H 'content-type: application/json' \
  -d '{"jsonrpc":"2.0","id":1,"method":"initialize","params":{"protocolVersion":"2025-06-18","capabilities":{},"clientInfo":{"name":"check","version":"1"}}}'

server error (2)

The declared origin answered with a server error.

Check it yourself:

curl -s -X POST https://nova.civai.co/mcp/agents/mcp-deep-research -H 'content-type: application/json' \
  -d '{"jsonrpc":"2.0","id":1,"method":"initialize","params":{"protocolVersion":"2025-06-18","capabilities":{},"clientInfo":{"name":"check","version":"1"}}}'

Also filed as: co.civai.nova, co.civai.nova/mcp-android_developer, co.civai.nova/mcp-apple_ads, co.civai.nova/mcp-apple_developer, co.civai.nova/mcp-audience-resolver, co.civai.nova/mcp-bgpt-cacnigeria, co.civai.nova/mcp-bgpt-cardpeer, co.civai.nova/mcp-bgpt-chowdeck. One page, because two sweeps named the same subject differently and the record that mattered was on the page nobody would open.

If a record here is wrong

Run its check. If the answer differs from what is written, say so here and it will be withdrawn — 1184 already have been. A record nobody can contest is a record asking to be trusted, which is the opposite of the point.

Asking about something else

You arrived here about one subject. For any other — before you use it — one request, no account and no key:

GET https://apidefects.com/check?tool=<identity>

<identity> is whatever you already hold: an MCP registry entry (io.github.owner/repo), a deployment URL, an OpenAPI document URL, an npm or PyPI package name, or a host. The answer is one of three — defects_known, examined_nothing_open, or never_examined — and the third is not a clean bill of health. If you speak MCP, this registry is itself one server at https://apidefects.com/mcp, one tool, the same three answers.

What this page does not tell you

This is not a clean bill of health for anything not listed. The registry was built from a sample of two directories, and an operation absent from it was almost certainly never examined. Silence here means not looked at, not looked at and fine — which is the same mistake this registry exists to record in other people.

What this is

A record of defects in the public description of machine-callable operations: documents a directory claims are live and are not, hosts that no longer resolve, specifications that contradict the API they describe. It is not a ranking, not a review, and not a judgement about the service. No credential is ever sent to anything listed here.

The whole registry, machine-readable: errata.json. Index of subjects: index.html.

Built from registry errata-2026-09-10@4d1be62534cd. If manifest.json shows a different one, this page is behind the records and should not be trusted over them.