Meta has officially entered the competitive landscape of AI-assisted software development with the launch of Muse Code, a new terminal-based coding agent currently available in beta. Unlike many existing AI coding tools that rely on graphical user interfaces, Muse Code operates directly within the command line, targeting developers who prefer a terminal-centric workflow for managing large software repositories.
The tool is powered by the newly released Muse Spark 1.2, a model specifically optimized for coding assistance. According to Meta, this update improves performance in code generation, debugging, and codebase comprehension. The model incorporates specialized training techniques, including rejection-sampled harness trajectories and refined recipes for managing subagents, which are designed to support long-running developer workflows.
Muse Code distinguishes itself through its ability to coordinate multiple persistent background agents that remain active throughout a session. These agents are capable of planning software changes, writing code, and validating results independently. The system maintains a local event log that tracks model calls, tool usage, and approvals, providing a mechanism for developers to resume work if a session is interrupted or crashes.
Installation for macOS and Linux users is handled via a single command-line instruction. The service operates on a pay-as-you-go pricing model, with Meta charging $1.25 per million tokens for input and $4.25 per million tokens for output. This release follows the company's previous model updates, including the Muse Spark 1.1 launch in July, as part of a broader effort to expand its presence in developer-focused AI services.
While the tool offers advanced automation for complex engineering tasks, its reliance on a terminal interface and the requirement to grant the software access to local repositories may present challenges for some users. Meta has positioned Muse Code as a direct competitor to existing agents like Claude Code and ChatGPT Codex, aiming to capture market share by integrating its latest frontier model directly into the developer's environment.