Meta launches Muse Code, a terminal AI agent for large code bases

Meta has been seen as a latecomer in the AI coding assistant race, but this week it made a notable move to close the gap. The company released Muse Code, a beta terminal-based coding agent designed for developers tackling big projects. In a Wednesday social media post, Mark Zuckerberg described it as capable of "complete software engineering tasks across large repos," spanning planning, code writing, and result validation.

The agent is powered by Meta's recently released coding model, Muse Spark. One of its key features is the ability to spawn sub-agents that work in parallel. According to Zuckerberg, for large jobs, Muse Code "fans out to separate sub-agents working in parallel in isolated worktrees," leaving the developer's working copy untouched. He cited internal testing where the agent built six features for a game simultaneously without collisions.

With this release, Meta positions itself as a cost-effective alternative to existing tools like OpenAI's Codex and Anthropic's Claude Code. Meta's AI chief, Alexandr Wang, told the Wall Street Journal that Muse Code "can be an incredibly good option, especially from a cost perspective." The company has been ramping up its AI investments, and this marks a step beyond its ad-focused AI work into developer-facing tools.

For developers wrestling with sprawling repositories, the promise of hands-off parallel execution is appealing—if it delivers. The key metrics will be reliability, cost per task, and how well it integrates into existing workflows. As with any new coding agent, the real test will be in messy, real-world codebases, not just demos.