Z.AI GLM Coding Plan Setup

This guide walks you through setting up Z.AI GLM Coding Plan as an LLM provider in Crucible.

What is Z.AI GLM Coding Plan?

Z.AI offers a GLM Coding Plan that provides two API endpoints for use with Crucible:

Native ZAI endpoint (recommended): An OpenAI-compatible API at https://api.z.ai/api/coding/paas/v4. You use GLM model names like GLM-4.7 directly. Configure with type = "zai".
Anthropic proxy endpoint (alternative): An Anthropic-compatible API at https://api.z.ai/api/anthropic. You send requests using Claude model names (e.g., claude-sonnet-4-20250514), and Z.AI maps them to equivalent GLM models. Configure with type = "anthropic".

Both endpoints use the same auth token via the x-api-key header.

Prerequisites

Crucible CLI installed
Z.AI GLM Coding Plan account with auth token

Setup

Step 1: Get Your Auth Token

Log in to your Z.AI account
Navigate to your API credentials or settings
Copy your GLM Coding Plan auth token (typically 49 characters)

Step 2: Set the Environment Variable

export GLM_AUTH_TOKEN="your-token-here"

Add this to your shell profile (~/.bashrc, ~/.zshrc, etc.) to persist across sessions.

Step 3: Configure the Provider

Add to your crucible.toml or llm_providers.toml.

Native ZAI (recommended):

[llm]
default = "zai-coding"

[llm.providers.zai-coding]
type = "zai"
endpoint = "https://api.z.ai/api/coding/paas/v4"
api_key = "{env:GLM_AUTH_TOKEN}"
default_model = "GLM-4.7"
temperature = 0.7
max_tokens = 4096

Anthropic proxy (alternative):

[llm.providers.zai-anthropic]
type = "anthropic"
endpoint = "https://api.z.ai/api/anthropic"
api_key = "{env:GLM_AUTH_TOKEN}"
default_model = "claude-sonnet-4-20250514"

The {env:GLM_AUTH_TOKEN} syntax reads the key from your environment variable at runtime, keeping secrets out of config files.

Important: Model Names

Which model names to use depends on which endpoint you configured:

Native ZAI (type = "zai"): Use GLM model names like GLM-4.7
Anthropic proxy (type = "anthropic"): Use Claude model names like claude-sonnet-4-20250514

Mixing them up will return an “Unknown Model” error. GLM names won’t work on the Anthropic proxy, and Claude names won’t work on the native endpoint.

Usage

Start a Chat

# Use default Z.AI provider
cru chat

# Specify provider explicitly
cru chat --provider zai-coding

Switch Models at Runtime

Inside a chat session, use the :model command to switch models without restarting:

:model GLM-4.7

Use model names that match your configured endpoint type.

Troubleshooting

”Missing API key”

Ensure GLM_AUTH_TOKEN is set in your environment:

echo $GLM_AUTH_TOKEN

If empty, set it and restart your terminal.

”Unknown Model”

You’re using a model name that doesn’t match your endpoint type.

Native ZAI (type = "zai"): use GLM-4.7, not claude-sonnet-4-20250514
Anthropic proxy (type = "anthropic"): use claude-sonnet-4-20250514, not glm-4-flash

Empty response / no output

Crucible surfaces this as: "LLM returned empty response — no content received from provider". The provider returned no content in its response.

Check:

Your endpoint URL is correct for your provider type
The model name matches the endpoint type (GLM names for native ZAI, Claude names for Anthropic proxy)
Your auth token is valid and hasn’t expired

Stream timeout

Error: "LLM stream timed out — no response within timeout period". The provider is reachable but isn’t responding within the timeout window.