Meta has introduced Muse Code (Beta), its first AI coding agent, alongside Muse Spark 1.2, an upgraded coding model designed to power the new platform, marking another major step in the company’s push to accelerate AI-driven software engineering.
The announcement was made by Meta Chief Executive Officer, Mark Zuckerberg, who said the latest releases demonstrate the company’s rapid pace of innovation across its artificial intelligence stack.
The launch comes just one month after the release of Muse Spark 1.1, underscoring Meta’s commitment to continuously improving its AI models and developer tools. Both Muse Code and Muse Spark 1.2 are now available to developers worldwide through expanded access on the Meta Model API.
Muse Code is a terminal-based AI coding agent that enables developers to execute complex software engineering tasks using natural-language instructions. Rather than simply generating code, the system can plan projects, write code, run tests, verify outputs, and manage entire code repositories with minimal human intervention.
A standout feature of Muse Code is its multi-agent architecture, which allows multiple AI agents to work in parallel on different aspects of a project, significantly reducing development time for complex engineering tasks.
Meta said the tool can be installed with a single command and will be offered under two pricing tiers, including an optional contributor plan designed to provide more affordable access for developers.
Complementing the coding agent is Muse Spark 1.2, an upgraded AI model co-trained with Muse Code to deliver tighter integration and improved coding performance. According to Meta, the new model offers higher first-attempt success rates, cleaner code execution, and reduces the need for repeated prompts.
Designed specifically for long-running, multi-file software development projects, Muse Spark 1.2 features a one-million-token context window, enabling it to process and understand significantly larger codebases than previous versions.
Meta described the model as highly competitive and said it represents another milestone in the company’s journey toward developing frontier AI models. The company also hinted that even larger and more capable AI models are currently in development.
The launch reinforces Meta’s growing ambitions in the rapidly evolving AI coding space, where technology companies are racing to build intelligent software development assistants capable of transforming how developers write, test, and maintain code.