Meta enters the coding-agent race with Muse Code

Meta has released Muse Code, a terminal-based AI coding agent now in beta for macOS and Linux. Powered by the company’s new Muse Spark 1.2 model, the tool is built to handle complete software-engineering tasks across large code repositories.

Muse Code can plan changes, write code, validate its work, and coordinate persistent background agents that stay active during a session. Its local event log records model calls, tool usage, approvals, and edits, allowing developers to resume work after a crash. Built-in commands can generate a plan, stress-test it, or continue working toward a defined goal.

“Releasing Muse Code in beta today. It’s a terminal coding agent that takes on complete software engineering tasks across large repos: planning changes, writing code, validating the results.”

Muse Code versus other coding agents

Meta says Muse Code differs from tools such as ChatGPT Codex and Claude Code in one obvious way: it currently has no dedicated app and runs through the terminal. Mac users can install it with a single command, while the source reports support for Linux but does not provide further installation details.

Muse Spark 1.2 is described as a coding-focused update with improvements in code generation, debugging, codebase understanding, and long-running developer workflows. It is available through Muse Code and the Meta Model API, which now has expanded global access.

The release builds on Meta’s Muse Spark 1.1 launch, which introduced the company’s paid API service and upgraded agentic and multimodal capabilities. Meta has not announced pricing for Muse Code in the available information, and there is no Windows version listed. For now, the product’s strongest distinction is its persistent-agent workflow—not a graphical interface or a disclosed cost advantage.

Recommended reading

Google names Gemini 4 boss as two co-leads leave