Meta takes on Anthropic and OpenAI with its first AI coding agent, Muse Code

Meta Platforms Inc. has unveiled its first AI coding agent, Muse Code, in an effort to compete with industry leaders Anthropic PBC and OpenAI PBC. Announced in a blog post, Muse Code is the creation of Meta's AI chief Alexandr Wang, who leads the company's Superintelligence Labs unit. The tool, now available in beta, can handle "complete software engineering tasks across large repos," according to Meta Chief Executive Mark Zuckerberg.

Muse Code operates as a terminal coding agent, taking on responsibilities such as planning changes, writing code, and validating results. The agent is powered by Muse Spark 1.2, an updated coding-focused model. Meta believes that coding has become one of the most successful applications of generative AI, with thousands of enterprises utilizing the technology to enhance their developer teams.

The company aims to make Muse Code more affordable than competing tools like Anthropic's Claude and OpenAI's Codex. Developers can access the tool through various pay-as-you-go tiers, with a "contributor tier" priced more than 10-times cheaper than other options. However, to use this affordable tier, developers must allow Meta to access their data.

Muse Code is underpinned by a harness that developers can use to manage different AI models tailored to specific coding projects. The AI agent is accessible through Meta's APIs and the OpenRouter platform, which hosts numerous "open-weight" models from Chinese AI labs such as DeepSeek Ltd. and Moonshot AI.