Meta announces Muse Code, a coding agent to rival OpenAI Codex and Claude Code

Meta has announced Muse Code, a new AI-powered coding agent for developers. The tool, which has been released in beta, is positioned in the same space as solutions like OpenAI Codex and Claude Code.

According to information shared by Meta CEO Mark Zuckerberg on his social media account, Muse Code can handle end-to-end software engineering tasks such as planning changes in large code repositories, writing code, and verifying the resulting outcomes.

HOW DOES MUSE CODE WORK?

The tool operates via the developer's terminal screen and can be installed with a single terminal command. Muse Code is built on the foundation of Muse Spark, the coding model previously announced by Meta.

According to Meta, the system can break down large-scale tasks into independent sub-agents when necessary. These sub-agents can work simultaneously in isolated environments to execute different coding tasks in parallel.

This mechanism is designed to perform operations without directly interfering with the developer's local working copy. This aims to reduce code conflicts and simultaneous development issues, which are particularly common in large teams.

In internal tests, it was stated that Muse Code developed six different features for a game project simultaneously without experiencing any code conflicts. Meta emphasizes that this feature will allow the tool to provide a more efficient workflow for large projects.

The performance of Muse Code during its beta phase indicates that competition in the AI-powered coding tools market will intensify further. Meta's new tool specifically targets teams working with large codebases.

News Source: 12punto