GLM-5.2 in Claude Code: setup with Jan AI and NVIDIA NIM

This tutorial connects GLM-5.2 to Claude Code through NVIDIA NIM, with Jan AI acting as a local proxy. The presenter explains that Claude Code uses Anthropic's API format while NIM provides an OpenAI-compatible endpoint, so the proxy translates between them. This setup uses a remote model endpoint rather than running inference on the viewer's hardware.

The demonstration uses Windows 11. The speaker also lists Mac and Linux availability for Claude Code and Jan AI. After installing Claude Code, the tutorial instructs viewers to log out of any Anthropic account. It then covers creating an NVIDIA account, verifying it with a phone number and generating an API key. The presenter describes the endpoint as free; the walkthrough does not establish unlimited access or usage limits.

In Jan AI, viewers enable NVIDIA NIM, paste the key and refresh the model list. The Claude Code integration maps the large, medium and small model slots to GLM-5.2. After selecting Save and Enable, the presenter keeps Jan's Local API Server running and reopens the terminal so Claude Code picks up the environment changes.

The final checks use /model, a short chat and a request to diagnose a web app in a project folder. The coding assistant begins inspecting and editing code, but the transcript does not confirm that it fixes the app. The presenter chooses Jan AI after reporting manual setup and troubleshooting difficulties with LiteLLM.