ZCode Setup Guide: Connect to the NextHub API
ZCode is the agent development environment (ADE) from Zhipu, and it supports adding any model service compatible with the Anthropic / OpenAI protocols as a custom provider. Once NextHub is configured, you can use Claude and GPT models directly inside ZCode.
Connection Steps
1. Install ZCode
Go to the ZCode website to download the Windows / macOS / Linux installer and complete the installation. On the welcome page at first launch you can skip the account connection and go straight to the main interface.
2. Open Model Settings
Click the model name in the input box → Manage Models (or open the Model Settings panel from Settings), then click Add Provider at the bottom of the provider list on the left.
3. Fill in the Provider Information (Claude Channel)
| Field | Value |
|---|---|
| Name | NextHub |
| API protocol | Anthropic |
| Endpoint (Base URL) | https://api.nexthub.cc |
| API Key | your NextHub ApiKey (a Claude group) |
Do not append /v1 to the Base URL
The endpoint for the Anthropic protocol follows the same convention as Claude Code's ANTHROPIC_BASE_URL: fill in the bare domain https://api.nexthub.cc, with no /v1 at the end. ZCode appends the Messages endpoint path automatically.
4. Add Models and Enable
Click Add Model and manually fill in the model IDs supported by this service, for example claude-opus-5-5 and claude-fable-5 (for the full list, see Channels and Models; the real-time data in the Model Plaza is authoritative). After saving, turn on the provider's enable switch.
5. Verify
Return to the conversation, select the model you just added in the model picker, and send a message. If you receive a reply, you are connected.
You Can Also Connect the GPT Channel
Add another provider: choose OpenAI for the API protocol, fill in https://api.nexthub.cc/v1 (with /v1) as the endpoint, use a key from the GPT-Azure (0.47x) or GPT-Pro (0.11x) group as the API Key, and set the model ID to gpt-5.6, gpt-6-astra, and so on. Both providers can coexist, and you can switch between them in the model picker at any time.
Troubleshooting
ZCode's provider configuration and the terminal environment variables are two independent entries that do not sync with each other. If the connection keeps spinning, first check whether the endpoint has an extra path — under the OpenAI protocol, ZCode appends /chat/completions automatically, and a redundant path in the address returns 404. The model ID must match the model list exactly.