Install
npx skills add MiniMax-AI/cli -y -gWho it is for
This skill fits when you want an agent or a terminal to generate text, images, video, or speech through official MiniMax capabilities. It is the official CLI of the MiniMax AI platform and is written as built for agents. The skill's main path is only one: add the skill to the agent. The install notes name OpenClaw, Cursor, and Claude Code, and also say “and so on”, so you must not read that as only those three. You need Node.js 18+ and a MiniMax Token Plan. If you have no Token Plan, or the environment is below Node 18, do not install yet.
It fits people who already chose MiniMax for content and only need an official skill entry. It does not fit people still looking for a fully free, zero-quota trial. The first run should generate one sentence of text, not a video or speech file.
First try in 3 minutes
Use only this official skill install command:
npx skills add MiniMax-AI/cli -y -g
Prerequisite: Node.js 18+ is available, and you already have a MiniMax Token Plan. The global and China subscribe pages are
https://platform.minimax.io/subscribe/token-plan
https://platform.minimaxi.com/subscribe/token-plan
The install location is the global skill your current agent reads. Do not put a terminal-global install command beside it.
After install, start a low-risk task: ask the agent to generate one MiniMax-only sentence of at most twenty characters, write no files, and generate no image, video, or speech. If it asks you to sign in, the interactive command in the material is mmx auth login; you can choose MiniMax OAuth or paste an API key. First confirm you are interacting on this machine. Do not put the secret into the chat or the repository.
Success signal: it returned one sentence produced through this skill path, and nothing was written to disk. If the skill does not appear, stop at the install command above. The material does not describe quota-spend details or a default model name.
Use cases
Case 1: Generate short text with the agent. Give a clear sentence request and ask for text only. Evidence covers text generation, but gives no model list, length cap, or price. Beyond that, stop at “can generate text”.
Case 2: Generate images, video, or speech. These three have larger side effects and files. First have the agent restate the type, topic, and path it will write, then run only after you confirm. Evidence lists these three next to text as official capabilities, but gives no resolution, duration, format, or save directory. Do not emit multiple videos in a row before that is confirmed.
Limits
The skill main path is only npx skills add MiniMax-AI/cli -y -g. The material also has a terminal-global install, but do not treat it as a replaceable main path beside skill install. Node.js 18+ and a Token Plan are required. The login command is authentication, not a second installer. A client roster must not be inferred as exclusive from the examples. The first try stops at short text; images, video, and speech all need a confirmed output location first. Do not invent model names or free tiers that do not appear. If you paste a key during auth, do it only inside the official interactive flow. Do not write it into skill files.