Meta Launches Muse Code for Large Software Projects

AI News

August 6, 2026

Meta Launches Muse Code for Large Software Projects

Meta has released Muse Code, a terminal-based AI coding agent designed to complete complex engineering tasks across large software repositories.

Available in beta for macOS and Linux, the tool can plan changes, write code and validate results. It is powered by Muse Spark 1.2, Meta’s latest coding-focused model.

Parallel coding agents

Muse Code can divide large jobs among persistent background agents working simultaneously in isolated worktrees.

This allows several features or fixes to be developed in parallel without modifying the developer’s active working copy. Meta says the approach reduces repeated information gathering and limits the amount of manual guidance required during longer tasks.

The agent also keeps a local event log of model calls, edits, approvals and tool activity. If a task is interrupted, it can resume from the same point instead of restarting.

Built for long-running tasks

Muse Spark 1.2 was trained for repository-wide development, complex debugging and extended workflows. Meta tested the system on GPU kernel optimization tasks involving more than 1,000 tool calls and sessions lasting up to 24 hours.

Muse Code places Meta in more direct competition with OpenAI Codex and Anthropic Claude Code as companies race to automate increasingly complex software engineering work.