Meta Launches Muse Code, an AI Coding Agent for Large Codebases

Takeaways

• Meta has launched Muse Code, a new AI coding agent for developers.
• It uses Muse Spark 1.2 to help plan, write, and validate code across large projects.
• Muse Code is available in beta for macOS and Linux.

Meta has unveiled Meta Muse Code, a new AI coding agent designed to help developers work more efficiently with large codebases.

Powered by the new Muse Spark 1.2 model, Meta Muse Code can plan code changes, write and validate code, and assist with complex software engineering tasks directly from the terminal. The company says the tool is built for repository-scale development rather than simple code completion.

Muse Code is launching in beta for macOS and Linux, joining a growing wave of AI-powered developer tools aimed at automating more of the software development workflow.

What Can Meta Muse Code Do?

Unlike traditional AI coding assistants that focus on autocomplete or answering programming questions, Meta Muse Code is designed to help developers work across entire software projects. It can understand large repositories, plan changes, write code, validate results, and keep track of long-running engineering tasks from within the terminal.

Muse Code can help developers:

• Explore and understand large codebases.
• Plan and implement multi-file code changes.
• Validate code before it's committed.
• Delegate tasks to background AI subagents.
• Recover ongoing work using a local event log if a session is interrupted.

Meta Muse Code is powered by Muse Spark 1.2, the company's latest coding model. It builds on Meta's earlier developer efforts, including the launch of the Muse Spark 1.1 Developer API, which first opened the Muse family to external developers.

Built for Large Codebases, Not Just Code Completion

Meta says Muse Code is designed for repository-scale software engineering rather than generating individual code snippets. Developers can assign broader tasks, review proposed changes, and let the agent work across multiple files while keeping track of progress through the terminal.

The company also introduced Muse Spark 1.2 alongside the coding agent, saying the new model was co-trained with Muse Code to better handle planning, reasoning, and long-running development tasks.

Muse Code is the latest addition to Meta's growing AI portfolio. Earlier this year, the company also introduced Muse, an AI image generator for creators and advertisers, showing how the Muse brand is expanding beyond image generation into developer tools.

Availability

Meta is releasing Muse Code in beta for macOS and Linux, with access initially focused on developers who want to experiment with AI-assisted software engineering. The company hasn't announced when a wider rollout or Windows support will be available.

Meta has been expanding the Muse platform at a rapid pace. We recently covered how Meta quickly rolled back one of Muse Image's Instagram features, showing how the company is continuing to refine its AI products as they evolve.

Why It Matters

AI coding assistants are evolving beyond code completion into tools that can understand entire software projects and help manage complex development tasks. With Muse Code, Meta is entering a growing market alongside AI-powered developer tools focused on improving productivity across large codebases.

While the beta is limited to macOS and Linux for now, Muse Code signals Meta's broader ambition to build AI tools that support professional developers across the software development lifecycle.