Meta Releases AI Coding Agent 'Muse Code' in Beta to Tackle Complex Engineering Tasks

Meta Platforms (NASDAQ: META) on Wednesday released a beta version of Muse Code, a new terminal-based coding agent designed to handle entire software engineering tasks from planning to execution across large code repositories.

The launch marks Meta's latest push into the competitive AI developer tools market, offering a solution aimed at automating complex coding workflows.

How Muse Code Works

Muse Code is powered by Muse Spark 1.2, an updated proprietary AI model specifically optimized for coding tasks. According to Meta, the agent is capable of planning changes, writing new code, and validating the results automatically.

A key architectural feature is its use of specialized background agents that remain active throughout a developer's session. This allows the tool to build and maintain context over time, unlike systems that restart with each new request. For larger tasks, Muse Code deploys parallel sub-agents that operate in isolated environments, or "worktrees," to avoid conflicts with the primary working copy of the code. The company reported that during testing, it successfully used this system to build six separate features for a game simultaneously.

Ad

Availability and Market Context

Meta stated that developers can install Muse Code with a single line of code and can begin using it through a "contributor tier." The company described the pricing for initial use as low-cost, signaling an intent to drive early adoption.

This release positions Meta more directly against established AI coding assistants from competitors like Microsoft's GitHub, Amazon, and Google. By focusing on automating complete, complex tasks rather than just code completion, Meta aims to capture a segment of the professional developer market seeking more advanced AI-driven productivity tools.