Meta has officially launched its AI programming agent, Muse Code, designed for large codebase development scenarios, helping programmers complete the entire software engineering process from task planning, code writing to result verification. The tool is currently in the testing phase. Meta's CEO Mark Zuckerberg stated that Muse Code can perform end-to-end development tasks in complex software projects, further enhancing AI's application capabilities in the engineering field.
According to the introduction, Muse Code is based on Meta's previously released coding model, Muse Spark, and can be installed and used through terminal commands. It adopts a multi-agent collaboration mechanism to handle large projects. When the task size is large, the system automatically splits the task, with multiple sub-agents executing in independent work trees in parallel, avoiding modifications to the user's current code environment. Zuckerberg mentioned that during testing, Muse Code successfully handled six features for a game development project without any code conflicts.
This release marks Meta's further commitment to the AI software development tools sector, aiming to enhance its competitiveness in the AI agent market. Currently, products such as OpenAI's Codex and Anthropic's Claude Code are already competing among developers. Meta hopes to attract more developers by leveraging lower costs and the advantages of a multi-agent architecture.
Alexandr Wang, Meta's head of artificial intelligence and director of the Meta Super Intelligence Lab, said that for a wide range of workflows and application scenarios, Muse Code could become a cost-effective choice. In recent years, Meta has continuously expanded its AI investments, not only using AI to optimize its advertising business but also launching an intelligent agent for customer service and support in June of this year, further positioning itself in the enterprise-level AI market. As AI agents evolve from code assistance tools to independently completing complex tasks, the competition for automation in the software development industry is accelerating.
