Meta, often seen as a latecomer in the AI assistant space, is now working hard to close the gap. This week, the company introduced a new terminal-based coding agent designed for developers who need help tackling complex tasks across extensive software codebases. The tool, called Muse Code, is currently in beta and can handle "complete software engineering tasks across large repos," according to a Wednesday social media post from Meta CEO Mark Zuckerberg. He noted that these tasks include "planning changes, writing code, validating the results."

Code, which can be installed with just a single command, runs on Meta’s previously launched coding model, Muse Spark. It manages large projects by spinning up its own agents that operate in parallel. "When a job is big enough, it fans out to separate sub-agents working in parallel in isolated worktrees," Zuckerberg explained. "Your working copy is never touched. In testing we had it build six features for a game simultaneously with no collisions."

This move appears to be an effort to position Meta as a stronger and more budget-friendly contender against AI Lab rivals like OpenAI, with its coding agent Codex, and Anthropic, which offers Claude Code. "We think that for a lot of workflows and a lot of use cases, this can be an incredibly good option, especially from a cost perspective," said Alexandr Wang, Meta’s AI chief who leads Meta Superintelligence Labs, in a conversation with the Wall Street Journal.

Meta has been ramping up its AI ambitions by investing heavily in development. In June, the company expanded beyond its traditional focus on using AI to boost its advertising business and stepped into the enterprise AI market with an agent tailored for customer service and support.