AsianFin — Chinese AI startup MiniMax has unveiled its latest general-purpose AI agent product, MiniMax Agent, designed to seamlessly integrate with mainstream tools and handle a wide range of multimodal tasks.

Powered by MiniMax’s proprietary MCP (Multimodal Control Platform) and embedded with commonly used industry tools such as Google Maps, GitHub/GitLab, Slack, and Figma, the MiniMax Agent is capable of “reading” long documents, “watching” videos, “listening” to audio, and “interpreting” images. It also features built-in generative capabilities for images, audio, and video content.

According to the company, MiniMax Agent can not only code complex webpages and web-based games—including those with intricate components and navigation logic—but also perform automated end-to-end testing by simulating user interactions, ensuring that the final outputs are stable and bug-free.