Docs · Providers

Huawei Cloud MaaS Token Plan in Claude Code and Codex

Huawei Cloud ModelArts MaaS's Token Plan is for personal accounts in 西南-贵阳一 (Guiyang) and serves GLM, Kimi and DeepSeek over Chat Completions and Anthropic Messages. Add Huawei Cloud in magpie with the plan, and Claude Code, Codex, Gemini CLI and OpenCode use them through magpie's local gateway on 127.0.0.1:3425.

Add Huawei Cloud

Add the provider and pick the planOn Providers, click ＋ Add provider, pick Huawei Cloud MaaS, set the key's plan to Token Plan or Pay as you go, paste the key and save.
Pick it for a toolOn Agents, pick a model such as glm-5.3 on a tool's row.

Terminal

magpie claude huaweicloud/glm-5.3
magpie codex huaweicloud/deepseek-v4.1-flash
magpie opencode huaweicloud/kimi-k2.6

Plans and endpoints

Plan	Chat Completions	Anthropic Messages
Token Plan	`https://api.modelarts-maas.com/plan/v2`	`https://api.modelarts-maas.com/plan/anthropic`
Pay as you go	`https://api.modelarts-maas.com/openai/v1`	`https://api.modelarts-maas.com/anthropic`

The Token Plan's quota is spent only at the endpoints under /plan, so magpie uses the plan's by default; pay as you go bills apart. Both take a MaaS API key, made under the ModelArts console's authentication management. Each key in the provider names its plan. Neither serves Responses.

Models

Until its list is fetched, magpie gives the plan's models: glm-5.3, glm-5.1, kimi-k2.6, deepseek-v4.1-flash, deepseek-v4-flash, lowercase as Huawei's OpenClaw page writes them. Once the list is in, Huawei's own list is the one served.

In each agent

Claude Code speaks Anthropic Messages, which the plan serves, so requests go as they are. See Claude Code with other models.
Codex speaks only OpenAI Responses, which Huawei Cloud doesn't serve, so magpie converts to Chat Completions; custom tools such as apply_patch go to the model as functions and come back as Codex can run them. See Codex with other models.
Gemini CLI speaks Gemini's API; magpie translates at the local gateway. See Gemini CLI with other models.
OpenCode: magpie writes the models into opencode.json. See OpenCode with other models.

FAQ

How do I use the Huawei Cloud Token Plan in Claude Code?

Add Huawei Cloud MaaS in magpie with the key's plan set to Token Plan, then pick glm-5.3 on Claude Code's row, or run magpie claude huaweicloud/glm-5.3. Requests go to the plan's Anthropic endpoint, https://api.modelarts-maas.com/plan/anthropic.

Does Huawei Cloud MaaS work in Codex?

Yes. Huawei Cloud has no Responses endpoint, so magpie converts Codex's requests to Chat Completions, custom tools included. Run magpie codex huaweicloud/deepseek-v4.1-flash.

Why the endpoints under /plan?

The Token Plan's quota is spent only at /plan/v2 and /plan/anthropic; requests to pay as you go's /openai/v1 bill apart. magpie uses the plan's endpoints by default.

Which models does the Huawei Cloud Token Plan have?

magpie gives glm-5.3, glm-5.1, kimi-k2.6, deepseek-v4.1-flash and deepseek-v4-flash until Huawei's list is fetched, and then serves that list.

The same for other providers: Tencent Cloud, Baidu Qianfan, Volcengine Ark, GLM Coding Plan, Kimi, DeepSeek, Qwen. Hit an error? See troubleshooting, or ask on Discord.