Meta is considered a bit of a laggard in the AI harness space, but is making strides to catch up. The company this week released a new terminal coding agent aimed at programmers looking for help with complex tasks across large software code bases.
Muse Code, currently in beta, can accomplish “complete software engineering tasks across large repositories,” Meta CEO Mark Zuckerberg said in a social media post Wednesday. These tasks include “planning changes, writing code, and validating results,” he added.
The code, which can be installed with a single command, leverages Meta’s previously released coding model, Muse Spark. Handle large projects by launching your own agents and working in parallel.
“If a job is large enough, it is deployed into separate subagents that run in parallel in independent work trees,” Zuckerberg explained. “We never touch the working copy. In testing, we built six features for the game at the same time without any conflicts.”
The move is aimed at positioning Meta more competitively and affordably with AI lab peers such as OpenAI and its coding agency Codex and Anthropic with Claude Code.
“We think this could be a very good option for many workflows and many use cases, especially from a cost perspective,” Alexander Wang, Meta’s head of AI who leads Meta Superintelligence Labs, told The Wall Street Journal.
Meta is trying to increase the presence of AI by pouring money into development. In June, it expanded beyond its core focus of using AI to support its advertising business and entered the enterprise AI market with agents for customer service and support.
