Meta Launches Muse Code, an AI Coding Agent to Rival Claude Code and OpenAI Codex

Meta has officially entered the AI coding agent arena. The company this week unveiled Muse Code, a new terminal-based coding agent powered by the updated Muse Spark 1.2 model, positioning itself as a direct competitor to Anthropic’s Claude Code and OpenAI’s Codex.

Asynchronous Agents for Complex Codebases

Muse Code, currently available in beta, can execute complex coding tasks across large repositories by launching multiple sub-agents that work in parallel within isolated worktrees. According to Meta CEO Mark Zuckerberg, the tool can handle “complete software engineering tasks across large repos,” including planning changes, writing code, and validating results.

Zuckerberg explained the tool’s architecture in a social media post: “When a job is big enough, it fans out to separate sub-agents working in parallel in isolated worktrees. Your working copy is never touched. In testing we had it build six features for a game simultaneously with no collisions.”

Cost-Competitive Strategy

Meta is considered a relative latecomer in the AI coding space. Anthropic’s Claude Code launched in 2025 as the first major agentic coding tool, with OpenAI’s Codex following shortly after. Meta’s strategy combines technical capability with a competitive pricing advantage.

“We think that for a lot of workflows and a lot of use cases, this can be an incredibly good option, especially from a cost perspective,” said Alexandr Wang, head of Meta Superintelligence Labs, in comments to the Wall Street Journal. Muse Spark 1.2 pricing remains consistent with the previous version at $1.25 per million input tokens and $4.25 per million output tokens.

Muse Spark 1.2: Purpose-Built for Code

The underlying model, Muse Spark 1.2, builds on last month’s Spark 1.1 release with coding-specific improvements for code generation, debugging, and codebase understanding. Developers can access the tool through Meta’s developer website or the OpenRouter platform using an API key.

Installation requires a single command, lowering the barrier for developers. However, given its beta status, the tool is currently best suited for software engineers and developers with technical backgrounds.

The Agentic Coding Race

The market for AI coding agents has intensified rapidly over the past year. These tools allow developers to describe requirements in natural language while AI autonomously handles code writing, testing, and deployment, fueling the rise of “vibe coding,” where even those with limited programming experience can build websites and applications.

Meta’s AI ambitions extend beyond coding tools. In June, the company expanded from its core advertising business into the enterprise AI market with agents aimed at customer service and support. The launch of Muse Code further strengthens Meta’s presence in the developer toolchain.

Source: TechCrunch · CNET