Meta released Muse Code, its first AI coding agent, on August 5, positioning the product as a direct competitor to OpenAI's Codex and Anthropic's Claude Code1,2,4.
Muse Code, currently available in beta, can accomplish complete software engineering tasks across large repositories, including planning changes, writing code, and validating results, Meta CEO Mark Zuckerberg said in a social media post. The agent is powered by Meta's previously released coding model, muse-spark, and can be installed with a single command.
The agent handles large projects by spawning sub-agents that work simultaneously in isolated worktrees. "When a job is big enough, it fans out to separate sub-agents working in parallel in isolated worktrees," Zuckerberg said. "Your working copy is never touched. In testing we had it build six features for a game simultaneously with no collisions".
Muse Code is the latest major release from Alexandr Wang, Meta's AI chief, who leads Meta Superintelligence Labs and oversees foundation model development. Wang joined Meta in June of last year as the centerpiece of Zuckerberg's effort to revamp the company's AI strategy.
Developers can access Muse Code via a pay-as-you-go option that Wang said contains similar API pricing as the Muse Spark 1.1 release, charging $1.25 per million input tokens and $4.25 per million output tokens.
"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," Wang said.
ANALYSIS The pricing disclosure signals Meta is competing on cost against OpenAI and Anthropic, consistent with its broader pattern of offering open or lower-priced AI tools.
The sub-agent architecture — parallel execution in isolated worktrees — represents a distinct design choice aimed at enterprise-scale codebases where merge conflicts and workspace contamination are persistent engineering concerns.