Chinese AI startup DeepSeek has introduced Super-Agent, an open-source prototype designed to make coding assistance more autonomous. The tool directly competes with Anthropic’s popular Claude Code.
Unlike conventional coding assistants, Super-Agent does not wait for a developer to type a request. It monitors group chats, identifies feedback on coding problems, and suggests possible solutions on its own.
Most AI coding tools follow a prompt-and-response model. A developer asks a question, submits code or describes a problem, and the assistant returns a solution. Super-Agent changes this workflow by treating developer conversations as a source of context, looking for signs of bugs, inefficiencies, or unresolved issues before a formal request is made.
DeepSeek’s Harness team announced the development on WeChat. The company informed, “The agent is built to take as little human attention as possible while delivering maximum value to developers. “
The approach could reduce the time engineers spend triaging problems and switching between tools. It also reflects a broader industry push to turn large language models into agents that can handle multi-step professional work.
Super-Agent comes from DeepSeek’s Harness team, formed earlier this year to develop AI systems that can compete with coding products from leading US AI firms, including Anthropic. It helps convert a code-generating model into an agent capable of broader software-development tasks.
DeepSeek rolled out the prototype on GitHub. However, its documentation describes Super-Agent as research software, rather than a fully commercial, production-ready product. The launch arrives as Chinese and American technology companies compete to build more capable and affordable AI tools.
Also Read: How AI Features May Affect Online Ad Privacy