Meta has released Muse Code, a terminal-based coding agent that starts at $1.25 per million input tokens, positioning price as its primary weapon against established competitors like Claude Code and Codex. The beta release runs on Muse Spark 1.2 and can plan, write, and test code changes across large repositories, according to a report from Channel Insider. The launch gives managed service providers and systems integrators a new option for AI-assisted development work, with cost serving as the company's most obvious edge.

Meta offers two pricing structures for Muse Spark 1.2: a Standard plan at $1.25 per million input tokens and $4.25 per million output tokens, where prompts and completions aren't used to improve Meta products, and a Contributor plan at $0.10 per million input tokens and $0.20 per million output tokens, where prompts and completions may be used to improve Meta products. The lower rates come with a trade-off that could push customer projects toward the Standard tier or require direct client sign-off, since the discounted pricing allows Meta to train on user data. Alexandr Wang, Meta's chief AI officer, told The Wall Street Journal that Muse Code could be "an incredibly good option" from a cost perspective.

The report states that Muse Code can maintain specialized background agents throughout a working session, allowing a primary agent to split larger tasks without repeatedly collecting the same repository data. A local activity log records model calls, tool runs, approvals, and edits, while built-in controls let developers approve or reject a plan before it runs. Interrupted tasks can pick up from the recorded state, and Muse Spark 1.2 was co-trained with Muse Code specifically for repository-scale generation and extended coding assignments. In one company test, the agent executed more than 1,000 tool calls across a run lasting up to 24 hours while optimizing GPU kernels.

The persistent agents and recovery features matter because channel providers have already watched AI agents enter DevOps workflows and security tools emerge for AI-written software, and Muse Code brings repository-scale assignments into the same deployment and review conversations that Claude Code and Codex already occupy. MSPs and systems integrators can help customers evaluate Muse Code, connect it to development environments, and set access rules for customer repositories. But Meta hasn't announced a reseller or managed-services program for Muse Code, even as OpenAI has launched a Partner Network and Anthropic has added service tiers and a partner hub for Claude deployments. Marketplace access, technical training, or a formal partner program could help providers turn Muse Code projects into repeatable services, and would signal whether Meta plans to compete for channel support as aggressively as it's competing on price. The strategic question for service providers centers on whether aggressive pricing alone can build the ecosystem partnerships that mature AI platforms already command, and whether clients will accept data contribution terms that fund those lower rates.