Meta's Muse Code: New AI Agent Challenges OpenAI & Anthropic

Meta has launched Muse Code, its first dedicated AI coding agent, now in beta. This terminal-based tool uses Muse Spark 1.2 to write, debug, and verify code. What's interesting is that Meta wants developers to use its AI for complete software-engineering jobs, not just for generating short pieces of code. Muse Code can plan changes, edit multiple files, and test software across large code repositories. It can even run several sub-agents simultaneously for complex projects. The agent also keeps a persistent activity log, which means if a session crashes, Muse Code can resume instead of starting over. Meta trained the model and agent together as a single system. The bottom line is Meta is directly challenging existing AI coding platforms like Anthropic’s Claude Code and OpenAI’s Codex, aiming to streamline software development.