Meta launches Muse Code, its first terminal coding agent

laptop screen displaying colorful code

Meta entered the coding agent market with the beta release of Muse Code, a terminal-based agent that competes directly with Anthropic’s Claude and OpenAI’s Codex.

What Muse Code Does

Install it with one command and it takes on full software engineering tasks: planning changes, writing code, and validating results across a wide range of use cases. It runs inside a single interface and can manage fleets of AI-powered agents to support the development process, according to Meta AI chief Alexandr Wang.

The agent ships alongside Muse Spark 1.2, the latest coding-focused model from Meta Superintelligence Labs. Unlike its predecessor Muse Spark 1.1, this version was developed and trained in parallel with Muse Code, which Wang says improves overall coding performance.

Pricing

  • Pay-as-you-go: $1.25 per million input tokens, $4.25 per million output tokens (same as Muse Spark 1.1 API pricing)
  • Contributor tier: More than 10 times cheaper than the pay-as-you-go rate, in exchange for opting in to help improve the model

The Operator Angle

If you’re already paying for Claude or Codex, Muse Code’s contributor tier pricing is worth watching. More than 10x cheaper than pay-as-you-go is a significant gap for high-volume use cases. The catch: you’re sharing usage data to help train the model, which is standard practice across the coding agent category.

Muse Code is currently available in beta. Wang leads Meta Superintelligence Labs, where he joined in June of last year as the centerpiece of Mark Zuckerberg’s effort to rebuild Meta’s AI strategy.

Stay on top of AI & Automation with BizStack Newsletter
BizStack  —  Entrepreneur’s Business Stack
Logo