Meta has entered the AI coding-agent race with Muse Code, a terminal-based tool built to handle complex engineering tasks across large software repositories. In beta and powered by the new Muse Spark 1.2 model, it fans big jobs out to parallel sub-agents. Meta is pitching it as a cheaper alternative to Anthropic’s Claude Code and OpenAI’s Codex.

Key Takeaways

• Meta launched Muse Code, a terminal AI coding agent in beta
• It handles complete tasks across large repos, planning to validating
• Big jobs fan out to sub-agents working in parallel
• It runs on Muse Spark 1.2, a coding-focused model update
• Meta positions it as a lower-cost rival to Claude Code and Codex

What Meta Launched

Meta, long seen as a laggard in AI coding tools, is moving to catch up. This week the company released Muse Code, described by CEO Mark Zuckerberg as a tool that can accomplish complete software engineering tasks across large repos, including planning changes, writing code, and validating results.

Key facts about the launch:

• Availability: In beta for macOS and Linux, installable with a single command
• Engine: Powered by Muse Spark 1.2, a coding-focused update to Muse Spark 1.1
• Model access: Muse Spark 1.2 is available worldwide via Muse Code and the Meta Model API
• Interface: A terminal-based tool aimed at developers already working in the command line

How It Handles Large Projects

Muse Code’s defining trick is parallelism. When a job is big enough, it fans out to separate sub-agents working in parallel in isolated worktrees, letting multiple features develop simultaneously without interfering with the main codebase.

It also adds persistence features:

• Async background agents that stay active throughout a session rather than launching per task, so the same context isn’t gathered repeatedly
• A local event log that records every model call, tool run, and operation, letting the agent resume exactly where it left off after a crash
• Bundled “skills” familiar from rivals: /plan turns a task into an approval-gated plan, /grill stress-tests that plan, and /goal drives toward a stated objective

Under the hood, Muse Spark 1.2 was co-trained with Muse Code to align the model and the agent, with Meta scaling up coding-task training compute and broadening training environments.

The Competitive Play

Meta is positioning Muse Code against the leaders on price. AI chief Alexandr Wang, who leads Meta Superintelligence Labs, told the Wall Street Journal the tool could be an incredibly good option, especially from a cost perspective.

The context for the move:

• It targets the same market as OpenAI’s Codex and Anthropic’s Claude Code
• It follows Meta’s June entry into enterprise AI with a customer-service agent
• Zuckerberg called Muse Spark 1.2 a step toward the frontier, with larger models coming

The Data Trade-Off

There’s a catch worth flagging. As VentureBeat noted, the default on-ramp for Muse Code sends developers’ code and prompts into Meta’s training pipeline, a trade-off enterprises with proprietary codebases must consciously opt out of by moving to standard pricing.

The details of that structure:

• The free contributor tier carries tighter rate limits (60 requests per minute versus 3,000)
• That tier is clearly aimed at individuals and experiments, not production workloads
• The approach is classically Meta: subsidize access, harvest data at scale, and use it to close the gap with the frontier

For developers legally required to keep code secure, that trade-off may not be one they can make.

Why It Matters

Muse Code is Meta’s most serious entry yet into a category it had largely watched from the sidelines. The launch shows the AI coding-agent race intensifying, with Meta competing on cost and scale rather than claiming outright superiority. But the data-harvesting default is the catch enterprises must weigh, and whether Muse Code wins developers will depend on real-world performance against entrenched rivals like Claude Code and Codex. For now, Meta has planted its flag in the fastest-moving corner of applied AI.

Digital Trendings is your trusted source for AI news and updates, stay tuned for more.