> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://docs.jambonz.org/guides/features/bring-your-own-llm/z-ai-glm/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.jambonz.org/_mcp/server.

# Z.ai (GLM)

[Z.ai](https://z.ai) (Zhipu AI) serves its GLM family of models behind an OpenAI-compatible API. The wire is OpenAI-compatible (chat completions, streaming, tools), so the integration is straightforward — jambonz talks to it via a base URL override.

Two endpoints are available:

* **Global** (default) — `https://api.z.ai/api/paas/v4`
* **China** — `https://open.bigmodel.cn/api/paas/v4`

## Get credentials

1. Sign in at the global console ([z.ai](https://z.ai)) or the China console ([open.bigmodel.cn](https://open.bigmodel.cn)).
2. Open **API Keys** and create a key.
3. Copy the value.

## Configure in jambonz

In the portal: **Account → LLM Services → + Add LLM Service → Z.ai**.

**`API Key`** `string` — required

The key from Z.ai's console.

---

**`Base URL`** `string`

Defaults to `https://api.z.ai/api/paas/v4` (global). Set to `https://open.bigmodel.cn/api/paas/v4` for the China (BigModel) endpoint, or to your own proxy.

---

Click **Test** to verify.

## Use in an agent verb

```js
session.agent({
  llm: {
    vendor: 'zai',
    model: 'glm-4.6',
    llmOptions: {
      systemPrompt: 'You are a helpful voice assistant.',
    },
  },
  stt: { vendor: 'deepgram', language: 'en-US' },
  tts: { vendor: 'cartesia', voice: 'sonic-english' },
  turnDetection: 'krisp',
  bargeIn: { enable: true },
  actionHook: '/agent-complete',
}).send();
```

## Available models

| Model id      | Notes                                                     |
| ------------- | --------------------------------------------------------- |
| `glm-4.6`     | GLM 4.6 — flagship, strong reasoning and tool use         |
| `glm-4.5`     | GLM 4.5                                                   |
| `glm-4.5-air` | Lighter/faster variant — good for low-latency voice turns |

Check the [Z.ai model catalog](https://docs.z.ai) for the live list.

## Quirks & errors

> **Note**
>
> **Global vs China**: keep `vendor: 'zai'` constant — only the `Base URL` (in the portal credential) differs between the global Z.ai endpoint and the China BigModel endpoint. Keys are not interchangeable across endpoints.

> **Warning**
>
> **Tool-calling capability varies by model.** GLM 4.6 handles tools well; lighter `-air` variants may be less consistent. Test thoroughly before relying on function calling.

> **Warning**
>
> **`401`** — key revoked, copied with whitespace, or used against the wrong region's endpoint. Regenerate in the console and confirm the `Base URL` matches.

> Configure jambonz to use Z.ai's GLM models over its OpenAI-compatible API.