Skip to content
InferenceHub

ZCode

Z.ai's desktop coding agent — the official harness for GLM-5.2 — on the Anthropic Messages wire. Reaches: Claude, plus the open-weight daily drivers; glm-5.2 is its home model. It supports custom providers natively, so it runs against InferenceHub with no config file — everything is entered in the GUI.

Install: download at zcode.z.ai.

60-second setup — adding InferenceHub as a custom provider in ZCode and running glm-5.2.
1

Add InferenceHub as a custom provider

In ZCode, open Settings → Model settings → Add provider, name it InferenceHub, and set the API format to Anthropic messages (/v1/messages). The Base URL is the bare origin — ZCode appends /v1/messages itself, so no /v1:

https://app.inferencehub.tech

Paste your key into the API key field:

sk-prov-live-YOUR_KEY
2

Add a model

Under Model list, add a model id exactly as the gateway lists it — glm-5.2 is the daily driver ZCode is built around, and any enabled Claude id works on the same provider:

glm-5.2
3

Use ZCode

In the chat box's model selector, pick the model under your InferenceHub provider — requests now route through InferenceHub, and your spend appears in the portal under Usage & Billing.

Troubleshooting

If you see…Fix
404 on every requestThe Base URL must be the bare origin https://app.inferencehub.tech — with /v1 on the end, ZCode posts to /v1/v1/messages.
"model not found"Add the id exactly as the gateway lists it — glm-5.2, not zai-glm5.2. List the live ids from the Models page.
Responses bill your GLM Coding Plan insteadYou picked the model under ZCode's built-in Z.ai provider. Choose it under the InferenceHub custom provider in the model selector.