Point any Claude Code at this server. It runs on the host's Claude subscription — no API key, no per-token bill.
Set Claude Code to talk to this server:
Run it:
claudePoint any OpenAI SDK / tool here. Models: any Claude id (runs on the sub) or glm-5.2 (1M context, runs on Cloudflare — separate):
Cline / Roo Code / Kilo Code: pick provider "OpenAI Compatible", set Base URL to , API key to the value above, and a model id (e.g. glm-5.2). Streaming, function-calling / tools, vision (Claude), and reasoning_effort are all supported.
Full tool-call example (works for glm-5.2 and Claude models):
Anthropic SDK / any tool — base URL + placeholder key:
Raw endpoint: