wtf is qiaomu-model-cli?

qiaomu-model-cli is a command line tool and agent skill that lets you call three different AI coding assistants installed on your own machine, Grok CLI, Kimi Code CLI, and Claude Code CLI, and run them at the same time instead of one after another. Instead of remembering the specific command line flags each tool needs, you describe a task and this tool routes it to the right assistant with sensible defaults already filled in. Each assistant has a role it defaults to. Grok is used for live web and X search, research, image generation, and video generation. Kimi, which supports a very large context window, is used for frontend work like UI, CSS, and component code. Claude Code is used for the most complex coding, refactoring, and code review tasks, and you can also explicitly pick a specific Claude model instead of the default. Tasks that do not depend on each other can be run as a batch, launching all three assistants at the same time and writing separate logs for each, which is faster than waiting for one to finish before starting the next. Tasks where one assistant needs the output of another, such as Grok researching a topic before Kimi builds a page from it, run in sequence instead through a separate mode. The tool streams each assistant's native progress events live, sends a periodic heartbeat so you can tell a still-running job apart from a stuck one, and writes full logs to disk with restricted file permissions since the raw output may contain task content. To avoid trusting a model's own claim that it finished, you can declare expected output files and verification commands for a job, and the tool checks those before marking a job successful. If a job fails, you can rerun just the failed jobs from a saved summary file instead of repeating the whole batch. Interrupting a run cleans up any background processes those assistants started. Using it requires the Grok, Kimi, and Claude command line tools already installed and logged in on your machine, along with Python 3.10 or newer. The project is released under the MIT License.