Meta Unveils Muse Code Beta Terminal Coding Tool

Meta has launched a new terminal-based coding agent named Muse Code in beta testing. The tool is designed to handle complete software engineering tasks within large code repositories, including planning changes and validating results. CEO Mark Zuckerberg announced the release on Wednesday via his post on X, confirming that the system operates using specialized background agents which maintain context over time rather than restarting from scratch for every task.

The agent features a local event log capability to ensure continuity if it encounters an error mid-task, allowing users to resume work seamlessly. Installation requires only a single line of code and offers a low-cost entry point known as the contributor tier. Zuckerberg noted that Muse Code is powered by Muse Spark 1.2, which represents the next step in their development towards frontier models with larger capabilities.

Meta first introduced Muse Spark in April, describing it as an AI model excelling at visual coding for creating websites and mini-games from prompts. This launch coincides with reported pressure on Meta to demonstrate how its artificial intelligence tools can generate revenue following significant investment spending. The company aims to provide meaningful growth justification alongside these financial demands.

Competitors are simultaneously advancing their own offerings in the AI coding sector. OpenAI revealed plans to expand Codex capabilities by acquiring Ona, a provider of secure cloud execution technology that allows delegation of long-running tasks without device restrictions. Additionally, reports indicate Google is discussing hiring talent and signing licensing agreements with startup Mechanize.

Earlier this year, news emerged regarding Google recruiting executives from Windsurf in July 2025 to bolster its artificial intelligence code generation capabilities. These developments highlight a broader industry trend where major technology firms are intensifying efforts to enhance their AI coding agents while navigating the challenges of monetisation and operational efficiency.