GLM (Zhipu) · mantis-agent-sdk

Zhipu's official international endpoint for GLM — glm-4.7 is their flagship agentic model; glm-4-flash is the fast/cheap tier.


endpoint	`https://api.z.ai/api/paas/v4`
env var	`ZHIPUAI_API_KEY` (aliases: `ZAI_API_KEY`, `ZHIPU_API_KEY`)
get a key	z.ai

Get an API key

Sign up at z.ai (mainland users: open.bigmodel.cn)
Open the API Keys page from your account menu
Click 'Create a new API key'
Copy the key now — it's shown once
Set ZHIPUAI_API_KEY (base URL api.z.ai/api/paas/v4)

Create a key ↗ · Pricing ↗

Free Flash-tier models; full GLM API is pay-per-token (trial credits for new users).

Enable

export ZHIPUAI_API_KEY=...        # shell profile — survives forever

or in the terminal — validates the key live before saving:

/enable glm

or pick any locked 🔒 GLM (Zhipu) model in /models and paste the key inline.

Models

Starter menu (once enabled, /models fetches the provider's full live list):

glm-4.7
glm-4.6
glm-4-plus
glm-4-flash

Switch anytime — /model glm-4.7 fuzzy-matches; context carries over.

SDK

from mantis_agent import MantisAgentOptions
options = MantisAgentOptions(
    model="glm-4.7",
    backend="https://api.z.ai/api/paas/v4",     # key read from $ZHIPUAI_API_KEY
)

Notes

ZAI_API_KEY and ZHIPU_API_KEY work as aliases.
glm-4.7 full-size is the model that's impractical to self-host — this endpoint is the sane way to run it.