Meta Launches Muse Code, an AI Terminal Agent for Complex Software Repositories
Meta has announced the launch of Muse Code, a new terminal-based AI coding agent available in beta. The tool is designed to assist programmers with complex software engineering tasks across large repositories by running parallel sub-agents in isolated environments.
What Happened
Meta Chief Executive Officer Mark Zuckerberg introduced Muse Code in a social media post on Wednesday. According to Zuckerberg, the terminal agent can carry out complete software engineering tasks across large code bases, including planning necessary changes, writing code, and validating results.
Installed using a single command, Muse Code is powered by Meta’s previously released coding model, Muse Spark. To manage extensive software projects, the system launches internal sub-agents that operate concurrently. Zuckerberg explained that when a task is large enough, the tool fans out to separate sub-agents working simultaneously in isolated worktrees, ensuring the main working copy remains untouched. During internal testing, the tool built six game features simultaneously without code collisions.
Key Highlights
- Beta Release: Muse Code is available in beta and installs with a single command.
- Powered by Muse Spark: The terminal agent relies on Meta’s previously launched Muse Spark coding model.
- Parallel Sub-Agents: Tasks fan out across sub-agents working concurrently in isolated worktrees, preventing conflicts in the main code copy.
- End-to-End Workflow: Handles planning, code writing, and result validation across large software repositories.
- Cost Focus: Meta leadership emphasized the agent’s competitive potential from a cost perspective.
Why This Matters
The release represents Meta’s latest attempt to compete with established AI coding tools, including OpenAI’s Codex and Anthropic’s Claude Code. Alexandr Wang, Meta’s AI chief and head of Meta Superintelligence Labs, told the Wall Street Journal that Muse Code offers an effective option for numerous workflows, particularly from a cost perspective. The move also aligns with Meta’s strategy to broaden its AI portfolio beyond advertising support, following its June entry into the enterprise AI market with a customer service agent.
What to Watch Next
As Muse Code undergoes beta testing, developer adoption and performance metrics across large software repositories will determine how effectively it competes with established terminal and coding agents in the enterprise space.
Frequently Asked Questions
What is Meta’s Muse Code?
Muse Code is a terminal-based AI coding agent currently in beta that automates software engineering tasks across large repositories.
How does Muse Code execute tasks across large code bases?
The tool deploys parallel sub-agents operating in isolated worktrees, allowing multiple tasks or features to be developed simultaneously without modifying the user’s primary working copy.
What underlying model powers Muse Code?
Muse Code is powered by Meta’s previously released coding model, Muse Spark.
Source: TechCrunch