
Meta launched Muse Code on August 5, an AI coding agent that the company says can handle complex tasks inside complex software. It extends Meta's existing AI coding offerings rather than starting from scratch, though Firstpost bills it as Meta's first coding agent proper — and positions it explicitly against OpenAI and Anthropic.
The large-code-base framing is the whole pitch. Meta has one of the biggest monorepos in the industry and internal tooling built around it, so "agent that works at repo scale" is the differentiator being claimed against assistants that shine on greenfield files and degrade across a million lines of legacy.
The fine print
What's not in either report: benchmark numbers, pricing, model backing, or who can actually use it today. Until Meta publishes specs, Muse Code is a positioning statement — coding agents are now table stakes for every frontier lab, and Meta was the conspicuous holdout.
Sources: TechCrunch, Firstpost