Meta Launches Muse Code, a New AI Coding Agent for Large Software Projects

Meta has released Muse Code, a new terminal-based AI coding agent designed to help developers manage complex tasks across large software repositories. Currently in beta, the tool can handle full software engineering workflows, including planning changes, writing code, and validating results, according to Meta CEO Mark Zuckerberg.

Powered by Meta’s previously released coding model, Muse Spark, Muse Code manages large projects by deploying multiple sub-agents that work simultaneously in isolated environments, leaving a developer’s original working copy untouched. Zuckerberg said the system was tested by building six separate features for a game at once without conflicts.

The launch marks Meta’s latest effort to compete more directly with rivals OpenAI, which offers the coding agent Codex, and Anthropic, maker of Claude Code. Alexandr Wang, who leads Meta Superintelligence Labs as the company’s AI chief, said the tool could prove especially attractive for cost-conscious developers across a range of workflows.

The release follows Meta’s broader push to expand its AI investments beyond advertising. In June, the company entered the enterprise AI market with an agent focused on automating customer service and support, as part of a wider strategy to grow its footprint across the AI industry.