Meta Launches Muse Code to Rival Claude Code and Codex

Mark Zuckerberg used a social media post on Wednesday to introduce Muse Code, Meta's new terminal-based AI coding agent, putting the company squarely into a market already dominated by OpenAI's Codex and Anthropic's Claude Code. The tool is now available in beta and runs on Muse Spark, a model built for handling large, sprawling codebases rather than isolated scripts.

What sets Muse Code apart, according to Meta, is how it manages scale. Instead of working through a single repository in sequence, it spins up its own sub-agents that operate in parallel across isolated worktrees. That design keeps a developer's active working copy untouched while multiple changes get built, tested, and validated simultaneously.

Meta says the agent can carry out full software engineering workflows on its own: mapping out what a change requires, writing the code, and then checking that the result actually works. In one internal test, Muse Code built six separate features for a game at the same time without any of them colliding, a demonstration Meta is using to argue the tool can handle real-world engineering complexity, not just toy examples.

Installation is deliberately simple. Meta says Muse Code can be set up with a single command, lowering the barrier for developers who want to try it against incumbent tools without a lengthy setup process.

Alexandr Wang, who leads Meta Superintelligence Labs as the company's AI chief, framed cost as one of Muse Code's main selling points. He said the agent could be an especially strong option from a pricing standpoint for a range of coding workflows, a pitch that directly targets teams weighing the ongoing token costs of running agentic coding tools at scale.

Muse Code could be an especially good option from a cost perspective for various workflows.

The launch extends a broader push by Meta into AI products aimed at businesses and developers rather than just consumers. In June 2026, the company entered the enterprise AI space with an agent built for customer service and support automation. Muse Code marks its next step, this time aimed squarely at engineering teams managing large, complex software projects.

By competing directly with Codex and Claude Code, Meta is signaling that coding agents have become one of the clearest battlegrounds in applied AI, where the value proposition increasingly comes down to how well a tool scales across large systems, and how much it costs to run it that way.