Q: How do I connect OpenClaw to 51codeX?
A: Use Custom Provider with OpenAI-compatible protocol. Set the base URL to https://www.51codex.com/codex/v1, use a 51codeX API key, and start with gpt-5.3-codex. For long sessions, set agents.defaults.compaction.reserveTokensFloor to 20000 or higher.
Q: What should I do if OpenClaw says "Context limit exceeded"?
A: That message usually comes from OpenClaw local compaction rather than the 51codeX API. Update ~/.openclaw/openclaw.json so agents.defaults.compaction.reserveTokensFloor is 20000 or higher, restart OpenClaw, and start a fresh conversation if the current thread is already very long.
Q: Why am I getting 401 or Unauthorized?
A: Make sure the API key was created on 51codeX and that you are using the correct base URL for Codex, Gemini, or Claude.
Q: What should I do if Gemini CLI says "fetch failed"?
A: Usually the environment variables were not applied correctly, or Gemini was launched from inside another Gemini session.
Q: Can I use Responses API through the Gemini channel?
A: Not at the moment. The current Gemini route supports Chat Completions, Messages, and Models, but not Responses.
Q: Why does the same model sometimes feel more expensive?
A: Long outputs, heavier reasoning, and larger total token volume will naturally consume more credits even within the same model family.
Q: Are 30-day plans calendar-month based?
A: No. Validity is counted from the activation timestamp. Daily quota reset at 00:00 CST is a separate rule.