Meta Seeks To Take on OpenAI and Anthropic With a New AI Agent For Software Developers

Meta is making a major push into AI-powered software development with the debut of Muse Code, its first dedicated coding agent, as the company seeks to compete with rivals OpenAI and Anthropic in one of artificial intelligence's fastest-growing markets.

The new tool, unveiled Wednesday, is the latest product from Meta Superintelligence Labs, the division led by AI chief Alexandr Wang. The launch represents another step in CEO Mark Zuckerberg's effort to transform Meta into a leader in advanced artificial intelligence while finding new revenue streams beyond its core advertising business.

"You can install it with one command and then use it to take on complete software engineering tasks across a wide variety of use cases, planning changes, writing code, validating the results," Wang said in an interview.

Unlike traditional AI assistants that primarily generate code snippets, Muse Code is designed to handle broader software engineering workflows. The agent can plan programming tasks, write and edit code, validate results, and help developers manage complex projects from a single interface.

The release comes as competition in AI coding tools intensifies. OpenAI offers Codex-powered coding capabilities, while Anthropic's Claude has become popular among developers for programming assistance. These products are increasingly being used to automate portions of the software development process, allowing engineers to focus on higher-level design and problem-solving.

Meta is positioning Muse Code as both a productivity tool and a more affordable alternative to competing services. Developers can access the preview version through a pay-as-you-go pricing model. Wang said the platform also includes a contributor tier that costs "more than 10 times cheaper" than the standard pay-as-you-go option, making price one of Meta's primary competitive advantages.

Rather than claiming superior performance over rival products, Wang said Meta is differentiating the Muse Spark family of AI models and Muse Code through lower costs. Muse Code is powered by Muse Spark 1.2, Meta's latest foundation model for coding tasks. According to Wang, the AI model and coding agent were developed and trained together, improving the system's overall software engineering capabilities.

While Wang declined to disclose adoption figures for Muse Spark, he said customer interest has been encouraging. "Adoption has been exciting and strong," he said. The coding agent operates through a software "harness" that enables developers to manage AI models specifically optimized for programming projects. Meta says the system is capable of coordinating multiple AI-powered agents that can assist with different stages of software development, from planning to testing.

Developers will be able to access Muse Code through the same Meta developer portal that hosts the company's Muse Spark application programming interface (API). Meta also plans to make Muse Spark 1.2 available through OpenRouter, a platform that gives developers access to AI models from companies including DeepSeek and Z.ai.

The company is also introducing a feature aimed at enterprise customers concerned about data privacy. Wang said Meta has begun accepting requests for zero-data retention, allowing developers to use Muse Code without the company retaining their data to improve future AI models. "It represents a big enterprise feature that is important for folks," Wang said.

The privacy option could help Meta appeal to businesses handling proprietary source code or sensitive internal software projects, an area where data security remains a major concern for AI adoption. Muse Code also represents an important commercial test for Meta's broader AI strategy.

The company has dramatically increased spending on artificial intelligence infrastructure, including new data centers and computing capacity, while reorganizing its AI efforts under Meta Superintelligence Labs. Those investments have drawn increased scrutiny from investors, particularly after Meta last week issued a softer-than-expected revenue forecast and reported declining free cash flow for the second quarter.

Because roughly 98% of Meta's revenue still comes from digital advertising, the company is under pressure to demonstrate that its AI investments can generate meaningful new sources of income.