AI API Compatibility & Test Evidence
Transparent status of Codex CLI, Claude Code, OpenAI Responses, Anthropic Messages and third-party endpoints. We only report verified results.
Current test status
| Test | Status | Evidence required to pass |
|---|---|---|
| Ofox OpenAI Responses API | Not verified | Successful authenticated response using a real funded key |
| Ofox Anthropic Messages API | Not verified | Successful messages response and supported tool calls |
| Codex CLI end-to-end | Not verified | Real coding task and streaming/tool-use results |
| Claude Code end-to-end | Not verified | Real client task and tool use |
| Cursor custom provider | Limited compatibility | Confirm capabilities individually; do not assume full native integration |
Earlier smoke tests were blocked by API credential formatting. A workflow exiting successfully without valid API responses does not count as a compatibility pass. No latency or success rates have been invented.
Review test workflow ↗Safe evaluation sequence
Confirm actual model availability and rates, fund only an appropriate small test balance, perform representative calls, test streaming and tools, and keep a working official-provider fallback. Never enter an API key on JXZNE.
Read English setup guides →