Meta AI has introduced Muse Code (beta), a terminal-based coding agent powered by its new Muse Spark 1.2 model. Positioned as a step toward the frontier of AI-assisted development, Muse Code is designed for complex software engineering tasks across large repositories. It handles planning, code writing, and validation, while asynchronous background agents persist throughout the entire session instead of being spawned per task. A local append-only event log records every model call, tool execution, approval, and edit, a feature Meta describes as "replay-exact" and "restart-safe." Notably, Muse Spark 1.2 was co-trained with the harness itself, and Meta has shared a kernel-optimization case study involving over 1,000 tool calls in a 24-hour period.
Deployability and Access
Muse Code is available in beta for macOS and Linux, installable via the command: curl -fsSL https://dev.meta.ai/install.sh | bash. Muse Spark 1.2 is accessible through both Muse Code and the Meta Model API, with expanded global availability. As of the 2026 release, no downloadable weights have been announced, so it functions as a hosted dependency.
- Company Level: Suitable for organizations of any size through the API; the Muse Code path is ideal for teams already using agent workflows with sandboxing and review gates.
- Industries: Best suited for software/SaaS, developer tooling, fintech engineering, GPU and inference infrastructure, semiconductors, and HPC.
- Applications: Repository-scale refactoring and migrations, long-running debugging tasks, and automated codebase maintenance.
via MarkTechPost